| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
Departments of
1
Pediatrics and
2
Pathology, The Johns Hopkins University School of Medicine, Baltimore, MD 21287.
3
The National Center for Health Statistics, Centers for
Disease Control and Prevention, Hyattsville, MD 20872.
a Address for correspondence: Blalock 1379, The Johns Hopkins Hospital, 600 North Wolfe St., Baltimore, MD 21287. Fax 410-955-3247.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
4
years, which included non-Hispanic whites, non-Hispanic blacks, and
Mexican-Americans. Apo B concentrations were the same in males and
females, lower in black males than in other males, low in childhood
(~0.80 g/L) and increasing to ~1.2 g/L in adults, and higher in
younger women on hormones. Apo AI was higher in females than males,
higher in blacks than in others, remained constant from childhood to
adulthood (~1.35 g/L) in males, but increased with age (~1.30 g/L
to ~1.55 g/L) in females, and was higher in women taking hormones.
These are the first national probability estimates of apo B and apo AI
in the US and are referable to the WHO-IFCC First International
Reference Materials for apo AI and B. |
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
By virtue of their associations with HDL and LDL, both apo AI and apo B reflect the relationships of the respective parent lipoproteins with risk for coronary heart disease (CHD) (6)(7). The circulating concentrations of both apo AI and HDL cholesterol are lower in patients with CHD than in those without serious disease. Conversely, both apo B and LDL concentrations are increased in patients with CHD. Numerous studies, which contain comparative information about the two apolipoproteins and their respective parent lipoproteins with respect to their associations with CHD, have been conducted (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).
Although the findings have been somewhat inconsistent, the majority of studies conducted over the past decade indicated that in univariate analyses, both apolipoproteins are at least as well correlated with the presence of CHD as the parent lipoproteins (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), and both low apo AI (32)(33)(35) and high apo B (32)(33)(34)(35)(36) appear to increase the risk for future coronary events.
These observations suggest that the two apolipoproteins may provide useful information with respect to assessing risk for CHD. To this end, Contois et al. (38)(39) recently reported the distributions of apo AI and apo B, primarily in adults, in the population-based Framingham Offspring Study with the use of methods that rely on the new WHO-IFCC First International Reference Materials for Apolipoproteins AI and B as the basis for accuracy. In the present study, we measured the concentration of the two apolipoproteins in Phase 1 of the Third National Health and Nutrition Examination Survey (NHANES III). This communication presents the distributions of apo AI and apo B in persons ages 4 years or older and is the first of such studies conducted on a national scale in the US population. As was true for the Framingham Offspring Study (38)(39), the apo AI and apo B distributions presented here also use the WHO-IFCC Reference Materials as the basis for accuracy.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Apo AI and apo B were measured in persons ages 4 years or older. Serum
samples were obtained by venipuncture from both fasting and nonfasting
persons examined in mobile examination centers and in the home.
Examinees were instructed to fast as follows. For individuals examined
in the morning, those ages 12–19 years were instructed to fast
8.5 h before examination; adults were instructed to fast 12 h
before examination. Subjects examined in the afternoon or evening were
instructed to fast 6 h. Children ages <12 years were instructed not
to fast. For this analysis, fasting subjects were defined as those who
reported fasting for 9 h or more. Nonfasting subjects were those
who reported fasting <9 h.
The NHANES III (1988–1991) population sample consisted of 16 919 persons ages 4 years or older. Of these, 14 269 were interviewed (84.3%) and 12 907 were interviewed and examined in mobile examination centers or at home, for an overall response rate of 76.3%. Apo B measurements were made in 11 483 subjects (67.9%), and apo AI in 11 432 subjects (67.6%). Selected characteristics known to affect serum lipoprotein concentrations (i.e., fat intake, alcohol consumption, use of exogenous hormones, smoking status, body mass index, physical activity) did not differ from those in subjects for whom apolipoprotein measurements were not made any more than would be expected by chance alone (data not shown). Approximately 6% of the adults in the study population as a whole and 5% of the female hormone users were taking ß blockers. Response rates were 72–73% in persons ages 20 years or older, 81–83% in those ages 12–19 years, and 85–86% in those ages 4–11 years. There were no differences in response rates for males and females in any age group.
Women were classified as hormone users if at the time blood samples were drawn they said they were taking birth control pills, taking estrogen or hormone pills, using vaginal cream or suppositories, receiving hormone injections, or using hormone patches (40). Otherwise they were classified as nonhormone users. Alcohol consumption was estimated from the self-reported number of times per day that beer, wine, and liquor were consumed over the 1-month period before examination (40). Physical activity was determined from the total number of times per day that the individual reported walking, running, or jogging, riding a mobile or stationary bicycle, swimming, doing aerobic exercise or aerobic dancing, doing other dancing, calisthenics, weight-lifting, gardening, or other forms of exercise during the 1-month period before examination (40).
laboratory methods
Blood sampling and shipment.
Blood was drawn from subjects in
the sitting position and allowed to clot for 45 min at room
temperature. An aliquot of the serum was transferred to a screw-capped
polypropylene storage vial and stored at -20 °C for up to 1 week
before being packed on solid CO2 and sent to the laboratory
by an overnight express delivery service. The samples were stored in
the laboratory at -20 °C for 1–2 weeks before analysis.
Apolipoprotein analysis.
Samples were thawed at room
temperature and mixed thoroughly for 30 min on a blood- rotating device
before analysis. The procedures used for apo AI and apo B analysis have
been described in detail (41). Briefly, apo AI and apo B
were measured by radial immunodiffusion (RID) in the first 8.2% (1055
specimens) of the specimens during the first 5 months of the study, and
by rate immunonephelometry (INA) for the remaining specimens during the
last 31 months. During the latter period, a 20% random subset of the
samples were also analyzed by RID to establish the relationships
between RID and INA (41). After confirming that the RID
method had remained stable between the initial 5-month period and the
subsequent 31-month period, the RID values for the first 5 months were
converted to equivalent INA values as described previously
(41). The apolipoprotein method was changed primarily
because it became evident shortly after the study began that automated
methods were becoming more widely used and we were unsure whether the
RID kits then being used (41) would continue to be
available for the entire study. We therefore decided to make this
change at the beginning and in such a way that would allow the RID and
INA data to be pooled and that could be adequately documented
(41).
At the beginning of the survey there were no standardized reference materials on which to base the measurements. Over the past few years, the WHO-IFCC First International Reference Materials for Apolipoproteins AI and B became available. The Northwest Lipid Research Laboratories, Seattle, WA, served as the coordinating laboratory for the development of these materials. For this reason, aliquots of the same random 20% subset of samples mentioned above were sent to that laboratory for analysis, and the results were used to transform the INA values to equivalent WHO-IFCC International Reference Materials-based values (41), which are presented here. The procedures used for QC have been described previously (41). The CVs for apo AI and apo B averaged <6% throughout the study.
lipid and lipoprotein measurements
Serum cholesterol and triglycerides and HDL and LDL cholesterol
were measured as described previously (42). The lipids
were measured enzymatically with the use of commercially available
reagents (Cholesterol/HP, cat. no. 816302, and Triglycerides/GPO, cat.
no. 816370, both from Boehringer Mannheim). HDL cholesterol was
measured in the clear supernatant after precipitating the other
lipoproteins with heparin and MnCl2 (1.3 g/L and 0.046
mol/L, respectively) and removing excess Mn2+ by
precipitation with NaHCO3, as described previously
(43). The biases (CVs) averaged -0.3% (1.7%), -2.1%
(3.9%), and 0.3% (3.4%) for cholesterol, triglycerides, and HDL,
respectively.
LDL was calculated from the three primary measurements with the use of
the Friedewald equation (44):
 | (1) |
statistical analyses
The apo AI and apo B results presented here were estimated with
the use of sampling weights. Each individual in whom apo AI and apo B
measurements were made was assigned a sample weight that corresponds to
the number of people in the US population represented by that
individual. The sample weights incorporate the differential
probabilities of selection, adjust for oversampling of subgroups of
interest, and include adjustments for nonresponse. The estimates of apo
AI and apo B presented here are based on weighted data
(40) and are representative of the noninstitutionalized
civilian population of the US. The reliability and precision of means
and percentiles were determined according to standard statistical
methods, as described previously (40). The minimum sample
sizes used for different percentiles were those recommended by NCHS
(40). The sample size depends on the percentile to be
estimated and design effects on the particular variable of interest,
y. Design effects measures of the impact of sampling design
on the variance of the estimates, expressed as the ratio
 | (2) |
of ref.
40, which summarizes minimum sample sizes according to
percentile and magnitude of the design effect.
|
All SEs were estimated, and all univariate and multivariate statistical analyses were performed with the SUDAAN statistical package (45)(46). The hypothesis that mean apo AI and apo B concentrations of hormone replacement users and nonusers were equal was tested at the 0.05 significance level by a Student's t-test statistic.
For the purpose of assessing the sensitivity and specificity of apo B
measurements as indicators of LDL cholesterol concentrations in the
increased-risk ranges of concentration, sensitivity was defined as the
proportions of subjects with LDL cholesterol concentrations 1.30- or
1.60 g/L who were correctly identified by virtue of having apo B
concentrations above cutoffs that were selected to maximize sensitivity
and specificity. These apo B concentrations were 1.07 g/L to detect LDL
cholesterol 1.30 g/L and 1.27 g/L to detect LDL cholesterol 1.60
g/L. Specificity was defined as the proportions of subjects with LDL
cholesterol concentrations <1.30 or <1.60 g/L who were correctly
classified by virtue of having apo B concentrations below the
respective apo B cutpoints. For apo AI, sensitivity was defined as the
proportions of subjects with HDL <0.35 g/L or with HDL 0.60 g/L who
were correctly identified by virtue of having apo AI concentrations
<1.14 or >1.54 g/L, respectively. These were the apo AI
concentrations that maximized sensitivity and specificity at the
respective HDL cholesterol cutpoints. Apo AI specificity was defined as
the proportion of subjects with HDL cholesterol 0.35 g/L or <0.60
g/L who had apo AI concentrations >1.14 g/L or <1.54 g/L,
respectively.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
. By design, 50% of the study subjects were males; 54% had
fasted at least 9 h before blood sampling.
apo b in fasting and nonfasting subjects
The National Cholesterol Education Program Adult Treatment Panel
II guidelines (47) recommend that LDL cholesterol
measurement be made only in patients that have fasted for 9 h. We
therefore compared apo B values in subjects who reported fasting 9
h or more with those in subjects who reported fasting <9 h. These
comparisons were made for all three race/ethnic groups stratified by
age and sex. There were no significant differences in apo B
concentrations in fasting and nonfasting subjects. We also examined apo
B concentrations in persons fasting <6 h compared with those fasting
9 h, stratified by sex, in all subjects and in the non-Hispanic
whites (data not shown). Again, significant differences were not
observed any more than would be expected by chance. Therefore, for the
following comparisons the data from fasting and nonfasting subjects
were pooled.
apo b by ethnic group
There was no significant difference in the age-adjusted mean
± SE apo B concentrations in all males (0.99 ± 0.01 g/L)
compared with all females (0.97 ± 0.01 g/L).
The age-adjusted mean apo B concentration for the three ethnic groups
are shown in Table 2
for children (ages 4–11 years), adolescents (ages 12–19
years), and adults (ages 20 years). Overall, the age-adjusted apo B
concentrations in males and females were similar; this was true in all
three ethnic groups. Black adult men had a slightly, but significantly
lower, age-adjusted mean apo B concentration than whites or
Mexican-Americans, whereas the means in females were about the same for
all ethnic groups. The concentrations remained stable throughout
childhood and adolescence, then increased in adulthood. This increase
was 37–38% in white and Mexican-American males and ~28% in black
males. For females, the increase was ~26% for all three ethnic
groups (Table 2
). The cumulative distributions of apo B in the three
ethnic groups are shown in Fig. 1
. These distributions for black males and females revealed lower
apo B concentrations than in whites or Mexican-Americans, but the
difference was greatest for the males (Fig. 1
).
|
|
apo b by age
Selected percentiles for apo B are shown in Table 3
. Median apo B concentrations in males were 0.75–0.79 g/L in
the age groups from 4 to 19 years. There was an abrupt increase in apo
B concentrations in adult men that continued from ages 20–50 years,
reached a plateau in those ages 50–69 years, and tended to decrease
after age 69 years (Fig. 2
). In females, the median apo B concentrations in the groups
ages 4–19 years were 0.79–0.82 g/L. Again, the concentrations
increased abruptly after age 20 years and reached a plateau after age
60 years (Fig. 2
). Because this was a cross-sectional study, it cannot
be determined whether the lower apo B values observed in the oldest
groups were due to a decline in apo B values or to selective mortality
in those with higher apo B values, although the latter factor almost
certainly operated. In both males and females, the concentrations of
LDL cholesterol also increased with age, and the increases were similar
to those observed with apo B (Fig. 2
).
|
|
The apo B pattern in women differed from that in men, however, in the
appearance of a double plateau, one that occurred in groups ages 20–49
years and a second that occurred after age 59 (Fig. 2
). The occurrence
of the first plateau in the younger women suggested the presence of
hormonal influences, possibly related to premenopausal hormone use,
postmenopausal changes, or both. For this reason, we examined apo B
concentration in adult women after stratification for hormone use. In
the group ages 20–49 years, hormone users had significantly higher apo
B concentrations than nonusers (Fig. 3
). After age 50, however, their apo B concentrations tended to
be slightly lower than in nonusers, but the differences were not
statistically significant (Fig. 3
). Furthermore, the double plateau
pattern apparent in the entire population was not seen in the nonusers
and had apparently resulted from hormone use in some of the younger
women.
|
We also examined LDL cholesterol concentrations in hormone users
and nonusers among adult women (Fig. 3
). The number of fasting hormone
users was insufficient to allow age stratification by decade, and wider
age strata were used. The LDL cholesterol concentrations in users and
nonusers were similar in the groups ages 20–39 years and 40–59 years,
but tended to be lower in hormone users after age 60 (P
= 0.062). Overall, LDL cholesterol increased ~19% between ages 20
and 60 in hormone users and ~35% in nonusers. This compared with an
~27% increase in apo B in users and a 42% increase in nonusers over
the same period. Age-adjusted mean apo B concentrations were
significantly higher in users than nonusers for the groups ages 20–49
years and significantly lower after age 50 (Table 4
). LDL cholesterol tended to be slightly lower in hormone users
overall (Table 4
), primarily because of the lower LDL cholesterol in
users over age 60 (Fig. 4
). As a consequence, the age-adjusted ratio of LDL cholesterol
to apo B concentrations was significantly lower in users than in
nonusers (P <0.001) (Table 4
). Triglycerides tended to be
slightly higher in hormone users.
|
|
Both male and female apo B concentrations ultimately reached about the same concentration (~1.20 g/L), but in males this occurred ~10 years earlier than in females. For the various age and sex strata, the differences between the mean and median values ranged from 0 to 0.05 g/L. This proximity of the means and medians to each other and the reasonably low skewness values for the various individual age and sex strata (range 0.02 to 1.40) and the skewness value for the entire population (0.74) suggested that the distribution of apo B was only slightly skewed.2
There was a reasonably high correlation between serum apo B and LDL
cholesterol (r = 0.87) (see also Fig. 2
), and the
correlation was the same in males and females. In all cases the mean
and median LDL cholesterol/apo B ratios were similar (data not shown).
As expected, there was slightly higher correlation (r =
0.92) between serum apo B and non-HDL cholesterol. Non-HDL cholesterol
is a measure of the amount of cholesterol associated with LDL plus the
other apo B-containing lipoproteins, i.e., VLDL, IDL, and Lp(a).
Because the LDL cholesterol in the NHANES III survey was estimated with
the use of the Friedewald equation (44), however, the
measurements included the contributions of both IDL and Lp(a)
(discussed in ref. 48). The slightly greater correlation
between apo B and non-HDL cholesterol was, therefore, due to the
contribution of VLDL to non-HDL cholesterol.
Ranges of apo B concentration were determined for adults ages 20 years
and older with LDL cholesterol concentrations in the various
risk-related ranges established by current National Cholesterol
Education Program guidelines for persons without CHD (47).
Table 5
shows the mean apo B concentration and the 5th through 95th
percentile ranges for subjects with desirable (<1.30 g/L), borderline
high (1.30–1.59 g/L), high (1.60–1.89 g/L), and very high (1.90
g/L) LDL cholesterol concentrations. There was a considerable overlap
in apo B concentration from one LDL cholesterol risk group to the next.
For example, the 5th–95th percentile range of apo B concentrations
observed in those with LDL cholesterol concentrations <1.30 g/L was
0.61–1.16 g/L, whereas that in subjects with LDL cholesterol
concentrations of 1.30–1.59 g/L was 0.94–1.38 g/L. Similarly, there
was a 0.26 g/L overlap between the groups with borderline and high LDL
cholesterol concentrations. Thus, it was not possible to use apo B
concentrations alone to classify subjects reliably into LDL risk
categories. On the basis of the findings in Table 5
, apo B measurements
would be expected to distinguish those with desirable LDL cholesterol
concentrations from those with high or very high concentrations, but
those with borderline LDL cholesterol concentrations could not be
distinguished from the other two groups reliably. This is perhaps not
too surprising given the variation in LDL composition that occurs among
individuals and the fact that apo B is found in all of the lipoproteins
except HDL. The extent to which apo B reflects LDL cholesterol depends
in part on the concentrations of the other apo B-containing
lipoproteins. For example, we found that only 1.8% of the subjects
with LDL cholesterol <1.30 g/L and triglycerides <2.00 g/L had apo B
concentrations >1.16 g/L, the 95th percentile for subjects with
desirable LDL cholesterol concentrations. In contrast, 27.4% of the
subjects with LDL cholesterol in the desirable range and triglycerides
of 2.00–4.00 g/L had apo B concentrations >1.16 g/L. The relative
contributions of apo B-containing lipoproteins other than LDL (as
opposed to LDL subfractions containing a lower than average ratio of
cholesterol to apo B) to the higher apo B concentrations in the latter
subgroup cannot be assessed from the present data.
|
Table 5
also shows the 25th through 75th percentiles for apo B
concentration in the various LDL cholesterol risk categories. With the
use of these percentiles, there was very little overlap in apo B
concentrations between risk categories, but these percentiles exclude a
substantial proportion of the population. Furthermore, 19.8% of the
subjects with desirable-range LDL cholesterol and triglycerides <2.00
mg/L had apo B concentrations >0.99 g/L, the 75th percentile for
individuals with LDL cholesterol <1.30 g/L. For those with desirable
LDL cholesterol and triglycerides of 2.00–4.00 g/L, 62.8% had apo
B concentrations >0.99 g/L.
We also examined the overall sensitivity and specificity of apo B for
identifying individuals with LDL cholesterol concentrations >1.30 and
>1.60 g/L, irrespective of triglyceride concentration. For this
purpose, we used apo B cutpoints that maximized both sensitivity and
specificity: 1.07 g/L for LDL 1.30 and 1.27 g/L for LDL 1.60 g/L.
Apo B had a sensitivity of 82.6% and a specificity of 85.6% for
classifying subjects with LDL 1.30 g/L, indicating that 14.4% of
subjects with desirable LDL cholesterol (i.e., ~8% of the entire
study population) would have been misclassified as having an LDL
cholesterol of 1.30 g/L or higher, whereas ~20% of subjects whose
LDL cholesterol actually exceeded this concentration (7.7% of the
total study population) would also been misclassified. Thus, 15.7% of
the of the total study population would have been misclassified. The
sensitivity and specificity of apo B in identifying subjects with LDL
cholesterol of 1.60 g/L or higher were 71.2% and 93.6%, respectively.
In this case, 28.8% of subjects with a high LDL cholesterol (5.3% of
the total population) and 6.4% of those with LDL <1.60 g/L (5.2% of
the total population) would have been misclassified, for an overall
misclassification rate of 10.5% with respect to the entire population.
apo ai in fasting and nonfasting subjects
The effect of fasting status on the estimates of apo AI
concentration was examined as discussed above for apo B. Again, no
significant differences were observed between fasting and
nonfasting subjects (data not shown). Therefore, the data from both
were pooled for the following analyses.
apo ai by ethnic group
The mean ± SE apo AI concentration in males (1.36 ±
0.01 g/L) was 0.10 g/L lower than in females (1.46 ± 0.01 g/L).
As seen from the age-adjusted apo AI values in Table 2
, the
male–female differences for all whites, ages 4 years or older (0.12
g/L), was twice that for all blacks (0.06 g/L). The difference in
Mexican-Americans (0.09 g/L) was intermediate between whites and
blacks.
The age-adjusted apo AI concentration in all black males (1.45 g/L) was
0.11 g/L higher than in white males (1.34 g/L), whereas that in
Mexican-American males (1.35 g/L) was essentially the same as in white
males (Table 2
). The apo AI concentration in black females was
significantly higher (1.51 g/L) than in white females (1.46 g/L) or
Mexican-American females (1.44 g/L) (Table 2
). The cumulative
distributions of apo AI for black males and females revealed higher apo
AI concentrations than in the other two ethnic groups, but as observed
for apo B, the difference was greatest in the males (Fig. 4
).
apo ai by age
Selected percentiles for apo AI are shown in Table 6
. Median apo AI concentrations in males remained essentially
constant with age and ranged from 1.28 to 1.36 g/L in all age ranges
except the group ages 6–11 years, which had a median value of 1.41 g/L
(Fig. 5
). The overall pattern of apo AI concentrations in females
differed from that in males because the median apo AI values tended to
increase with age, then reached a plateau in the groups ages 50 years
(Fig. 5
). There appeared to be somewhat less concordance between apo AI
and HDL cholesterol than between apo B and LDL cholesterol in both
males and females (Fig. 5
). Thus, whereas apo AI tended to remain
rather constant with age in males and to increase with age in females,
there appeared to be a slight downward trend in HDL in both sexes (0.02
to 0.03 g/L) in the groups with ages between 20 and 69 years (Fig. 5
).
Thecorrelation between apo AI and HDL cholesterol was 0.74 in males and
0.78 in females.
|
|
There appeared to be a fairly pronounced double plateau in women. The
first occurred in the groups ages 20–39 years, and the second after
age 50 (Fig. 5
). This double plateau thus occurred over the same age
ranges as for apo B in women, again suggesting the possible influence
of hormone use or hormonal changes that affect apo AI concentrations in
women. Unlike the pattern observed with apo B, however, stratification
of the adult women according to hormone use revealed that the double
plateau persisted in the nonhormone users, the first occurring in the
groups ages 20–49 years and the second after age 50 years (Fig. 6
). Hormone users had uniformly higher apo AI concentrations than
nonusers at all ages, and significantly higher HDL cholesterol
concentrations in the groups ages 40–69 years. In addition, the
age-adjusted concentrations of both apo AI and HDL cholesterol were
significantly higher in hormone users. The age-adjusted ratios of apo
AI to HDL cholesterol in hormone users and nonusers were virtually
identical (Table 4
), however, indicating that apo AI and HDL
cholesterol were affected similarly by hormones.
|
In all the age subgroups of males and females, the mean apo AI
concentration ranged from 0.01 to 0.05 g/L higher than the median
(Table 6
). The skewness values for the various sex and age strata
ranged from 0.24 to 1.28 and that for the entire population was 0.95.
As for apo B, the proximity of the means and medians to each other and
the reasonably low skewness values2
for the various
strata suggested that the distributions were slightly skewed.
To determine the sensitivity and specificity of apo AI for classifying
individuals with low (<0.35 g/L) or high (0.60 g/L) HDL cholesterol,
we used apo AI cutpoints that maximized sensitivity and specificity:
1.14 g/L for low HDL cholesterol and 1.54 g/L for high HDL cholesterol.
For low HDL, the sensitivity was 52.1%, and the specificity was
95.1%. Thus, apo AI misclassified 47.9% of the subjects with low HDL
(5% of the total population), and 4.9% of those with HDL cholesterol
0.35 g/L (4% of the population), for an overall misclassification
rate of ~9% with respect to the total population. Apo AI was
somewhat better at classifying subjects with high HDL cholesterol, for
which the sensitivity and specificity were 70.0% and 89.0%,
respectively. Thus 30% of individuals with HDL cholesterol 0.60 g/L
(8% of the population) and 11% of those with HDL cholesterol <0.60
g/L (8% of the population) would be misclassified, resulting in a
misclassification rate of 16% in the total population.
correlates of apo ai and apo b concentrations
To control for the confounding effects of various possible
correlates on apo AI and apo B concentrations, we performed multiple
linear regression analyses with either apo AI or apo B as the dependent
variable (Table 7
). The independent variables in the regression models included
hormone use, age, sex, race, body mass index, alcohol consumption,
cigarette smoking, physical activity, and years of education. The
strong and direct association between hormone use and apo AI
concentrations in females persisted after controlling the other
variables (P <0.001). The analysis indicated that in female
hormone users, the apo AI concentrations were 0.168 g/L higher than in
others. As expected, age, female sex, black race, alcohol consumption,
and years of education were independently and directly associated with
apo AI. Body mass index and current cigarette smoking were
independently and inversely associated with apo AI concentration.
Female sex, alcohol use, and hormone use were associated with larger
effects on apo AI, whereas body mass index, age, black race, current
smoking, and higher education were associated with smaller effects
(Table 7
).
|
For the most part, the opposite relationships were observed between
these correlates and apo B concentrations (Table 7
). Female sex,
hormone use, and black race were independently and inversely related to
apo B concentrations, whereas body mass index and current cigarette
smoking were independently and directly related to apo B
concentrations. Past cigarette smoking was also independently related
to apo B concentration (P <0.05), but not to apo AI
concentration; years of education were independently related to apo AI
(P <0.05) but not to apo B concentration.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The mean apo AI concentration in all males was ~7% lower than in females, reflecting male–female differences in HDL cholesterol concentrations (42)(50). The male–female difference in whites, however, was ~1.5-fold greater than in the other two ethnic groups. Overall, apo AI concentrations in men tended to remain constant with age.
In females, the age-related pattern was markedly different. Apo AI concentrations were lowest in childhood and adolescence and tended to increase continually through adulthood. As observed for apo B, apo AI in females also tended to plateau in the groups ages 20–40 years, then increased to a maximum concentration of 1.54 g/L thereafter. Unlike apo B, however, this remained true in both the hormone- and non-hormone-using women, indicating that the more gradual increase in apo AI with age in premenopausal women was not related to the use of exogenous hormones. Apo AI concentrations were higher in adult women using hormones than in nonusers at all ages, but HDL cholesterol concentrations were significantly higher in hormone users only after age 40. Again, the pattern of higher HDL cholesterol in older female hormone users has been observed by others (49).
The differences observed between either apo AI or apo B concentrations in subjects that had fasted at least 9 h compared with those who had fasted for shorter periods were no greater than expected by chance alone. About 16% of the subjects had fasted <6 h and 30% reported their last food intake to be 6–8 h before venipuncture. Although one cannot conclude with certainty that some postprandial changes may not have occurred in the 2–4-h postprandial period, when the increase in plasma triglycerides and the appearance of partially metabolized intestinal lipoproteins would be maximal, it is reasonable to infer that if such changes occurred they would have been minor, and the results suggested that both apolipoproteins can be measured in either fasting or nonfasting individuals.
In view of the variety of methods used for apolipoprotein analysis, the previous lack of common reference materials to which the calibration of the methods could be referred, and the different populations in which these measurements have been made (9)(20)(22)(27)(31)(51)(52)(53)(54)(55)(56)(57)(58)(59)(60)(61)(62)(63), it is difficult to compare the apo AI and B values reported in different studies (48)(53). However, the age-related changes in apolipoprotein concentrations as reported in several earlier studies can be compared. Kottke et al. (57) reported that apo B concentrations increased with age in males and females in the Rochester Family Heart Study. Similar observations were reported by Schaefer et al. (31) in the Framingham Offspring Study, Vermaak et al. (54) in South Africa, Kinlay et al. (55) in Australia, Zunic et al. (60) in healthy blood bank volunteers in Yugoslavia, and Brown et al. (62) in the Atherosclerosis Risk in Communities Study. The age-related increases were most prominent in women (31)(55). On the other hand, Sarihyan et al. (56) found little change in apo AI with age in either males or females. Brown et al. (62) also found no change in apo AI with age in the Atherosclerosis Risk in Communities Study, although the changes we observed in the present study occurred primarily at younger ages than those they studied (62). There also appeared to be only a slight tendency for apo AI and apo B to increase with age in a subset of the Lipid Research Clinics population in Seattle (51)(52), whereas Noma et al. (53) reported an age-associated decrease in apo AI, but apo B increased with age in both males and females. The reasons for the apparently conflicting findings are not immediately apparent, although changes in apolipoprotein concentrations with age may well be different in different populations. The results of such studies must be interpreted cautiously, however, because most were conducted in much smaller populations or over a more restricted age range than we examined. In the present study, there was a clear increase in apo B concentration with age in both males and females, and the increase in non-hormone-using women occurred at least 10 years later than in men. Furthermore, the patterns of age-related changes in apo AI concentrations differed in males and females; apo AI concentrations tended to increase only in females and, as observed by others (62), were higher in hormone-treated women.
In most cases serum apo B reflects primarily the apo B in LDL. In
addition apo B appears to be at least as good a discriminator of CHD as
LDL cholesterol. The question therefore arose whether apo B
measurements might be used reliably as a surrogate for LDL cholesterol
measurements to classify subjects according to National Cholesterol
Education Program risk amounts for LDL cholesterol. However, with an
apo B cutpoint of 1.54 g/L, which was selected to maximize the
sensitivity and specificity of the test, the sensitivity of apo B for
identifying individuals with high LDL cholesterol, i.e., >1.60 g/L,
was only ~70%; thus the measurement would be expected to misclassify
~30% of those with high LDL cholesterol, a misclassification rate
that in our opinion is unacceptably high. However, apo B might be used
to classify at least some individuals. On the basis of the findings in
Table 5
, apo B values <0.61 g/L or >1.57 g/L would reliably identify
individuals with LDL cholesterol <1.30 g/L and 1.60 g/L,
respectively. Within the range 0.61–1.57 g/L, it would be difficult to
distinguish between desirable, borderline high, and high risk
classifications. Furthermore, the misclassification of individuals with
desirable LDL cholesterol concentrations would be considerably greater
in those with triglyceride concentrations of 2.00–4.00 g/L, compared
with those with lower triglyceride concentrations.
For several reasons, these findings are perhaps not too surprising. First, each of the apo B-containing particles [VLDL, IDL, LDL, and Lp(a)] contains one molecule of apo B per particle, and all except VLDL are potentially atherogenic. Apo B, therefore, can be considered a measure of the number of potentially atherogenic particles present. Second, LDL cholesterol is usually calculated from measurements of total cholesterol, triglycerides, and HDL cholesterol (44) and was estimated in this way in NHANES III (42)(50). However, these LDL cholesterol measurements also contain the contributions of cholesterol from IDL and Lp(a). On average, these lipoproteins contribute ~0.02–0.04 g/L to the LDL cholesterol measurement, but can contribute more in individual patients. Furthermore, the cholesterol content of each of the lipoproteins can vary somewhat among individuals. Thus, LDL cholesterol reflects several characteristics of several lipoproteins: the relative concentrations of LDL, IDL, and Lp(a) within an individual; the different cholesterol compositions of these three lipoproteins within an individual; and the various cholesterol compositions of each of these lipoproteins among individuals.
The question of whether apo B might eventually be used instead of LDL cholesterol as the basis for assessing risk for CAD remains open. Several recent studies suggest that an increased apo B increases the risk for future coronary events (32)(33)(34)(35)(36)(37), but much further work is required to determine whether apo B satisfies accepted epidemiological criteria for establishing causality between high apo B concentration and CAD. Furthermore, the question remains whether the total number of apo B-containing particles present (as reflected by the apo B concentration), the concentration of cholesterol in these particles (as reflected by the LDL cholesterol measurement), or both considered together would be the more sensitive indicator(s) of risk for CAD.
Similar considerations apply for apo AI. It seems clear, however, that the lower correlations observed between apo AI and HDL cholesterol than between apo B and LDL cholesterol and the low sensitivity of apo AI for identifying patients with low HDL cholesterol concentrations would make it even more difficult to use apo AI as a surrogate measure of HDL cholesterol. Again, further research is required to determine whether the measurement of apo AI itself may provide a more sensitive indication of risk for CHD than HDL cholesterol currently provides.
Two final points should be considered that deal with the nature of the samples used for the NHANES III survey. First, by design, specimens collected in NHANES surveys were frozen and stored briefly before and after shipment to the laboratory (see Methods). While the possible effects of freezing on the apolipoprotein measurements were not evaluated in the present study, in several published studies freezing for periods of 1 month to 3 years at -20 °C or lower did not significantly change apo AI values (38)(51)(64)(65). For apo B, several investigators reported that frozen storage at -20 °C for periods ranging from 1 week to 6 months (66)(67)(68) or for 1 month at -80 °C (39) had no substantial effect on apo B values. On the other hand, Kafonek et al. (69) found a slight (6.5%) decrease in apo B values in >200 samples stored at -70 °C for almost 2 years. Brown et al. (70), reporting a similar decrease in 20 specimens stored at -70 °C for 6 weeks, attributed the decrease to the acute effect of freezing rather than to storage-related losses. In a recent examination of the acute effects of freezing on apo B measurement in >100 samples, however, we observed no detectable change when the samples were analyzed 24 h after freezing (unpublished observations). Considered together, the various studies suggest that the use of fresh-frozen samples probably did not affect apo AI values, and if freezing, per se, affects apo B at all, the effect in the present study was probably minor.
Additional, albeit indirect, support for these conclusions is also evident when comparing the present apo AI measurements with those measured by Contois et al. (38) in the Framingham Offspring Study. In both studies the apolipoprotein measurements are based on WHO-IFCC standards, which allows such a comparison. The mean apo AI values by age decade for males and females in the age decades common to both studies (i.e., adults ages 30 years or older) were almost identical, despite apo AI measurements in the Framingham Offspring Study being made in samples that had been stored for 4–6 years at -80 °C (39). A similar comparison for apo B, however, again with the use of the findings of Contois et al. (39) shows that the mean apo B values by decade averaged ~10% lower in the Framingham Offspring Study than in NHANES III. The basis for this difference is not certain. It may reflect actual differences in the two study populations, but might also have resulted in part from a fairly slow deterioration of apo B during the 4–6-year period that the Framingham Offspring Study samples were stored before analysis (39).
The second issue is the possible effect of sample turbidity, which generally occurs in samples with triglycerides >4.00 g/L and can influence immunonephelometric and immunoturbidimetric methods. For several reasons, it is unlikely that sample turbidity affected the results reported here. First, the NHANES III, Phase 1 subjects were selected to represent a normal population. The average triglyceride, by decade, ranged from 1.00 to 2.00 g/L (42), and there were only 60 individuals (0.5% of the Phase I population) with triglycerides exceeding 4.00 g/L. Second, detergent-containing buffers were used in the analytical methods to minimize turbidity and optimize immunoreactivity. Third, the instrumentation we used flagged samples that were too turbid to be analyzed directly and that had to be diluted for analysis. Finally, as we noted above, apo AI and apo B concentrations were the same in fasting and nonfasting subjects, suggesting that sample turbidity would not have markedly influenced the measurements in the NHANES III survey.
To our knowledge, this is the first national survey of apo AI or apo B distributions in the entire US population and the second US study in which the measurements can be referred to the new WHO-IFCC International Reference Materials for apo AI and apo B. The availability of these materials should now make it possible to compare apo AI and apo B concentrations in different studies and populations and ultimately lead to the development of common, risk-based cutoffs for these apolipoproteins similar to those currently used for LDL and HDL (47).
|
Acknowledgments |
|---|
|
Footnotes |
|---|
2 For comparison, preliminary examination of Lp(a) values in NHANES III yielded skewness values of 2.0 or higher. Lp(a) distributions are known to be highly skewed in most populations. The skewness value for an unskewed distribution would be 0.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
