Hazard ratios (HRs) are adjusted for age (years), blood pressure (Framingham
categories), diabetes, current smoking status, and body mass index. CI indicates
confidence interval; and CRP, C-reactive protein.
According to increasing quintiles of each measured variable over the
10-year follow-up period.
Hazard ratios are adjusted for age (years), blood pressure (Framingham
categories), diabetes, current smoking status, and body mass index. CRP indicates
C-reactive protein; HDL, high-density lipoprotein. Similar results were observed
for total cholesterol and low-density lipoprotein cholesterol and the ratios
of low-density lipoprotein cholesterol to non–HDL cholesterol and apolipoprotein
B100 to HDL cholesterol (data not shown). Error bars indicate 95%
Customize your JAMA Network experience by selecting one or more topics from the list below.
Ridker PM, Rifai N, Cook NR, Bradwin G, Buring JE. Non–HDL Cholesterol, Apolipoproteins A-I and B100, Standard Lipid Measures, Lipid Ratios, and CRP as Risk Factors for Cardiovascular Disease in Women. JAMA. 2005;294(3):326–333. doi:10.1001/jama.294.3.326
Context Current guidelines for cardiovascular risk detection are controversial
with regard to the clinical utility of different lipid measures, non–high-density
lipoprotein cholesterol (non–HDL-C), lipid ratios, apolipoproteins,
and C-reactive protein (CRP).
Objective To directly compare the clinical utility of total cholesterol, low-density
lipoprotein cholesterol (LDL-C), HDL-C, non–HDL-C, apolipoproteins A-I
and B100, high-sensitivity CRP, and the ratios of total cholesterol
to HDL-C, LDL-C to HDL-C, apolipoprotein B100 to apolipoprotein
A-I, and apolipoprotein B100 to HDL-C as predictors of future cardiovascular
events in women.
Design, Setting, and Participants Prospective cohort study of 15 632 initially healthy US women aged
45 years or older (interquartile range, 48-59 years) who were enrolled between
November 1992 and July 1995. All participants were followed up over a 10-year
period for the occurrence of future cardiovascular events.
Main Outcome Measure Hazard ratios (HRs) and 95% confidence intervals (CIs) for first-ever
major cardiovascular events (N = 464) according to baseline levels
of each biomarker.
Results After adjustment for age, smoking status, blood pressure, diabetes,
and body mass index, the HRs for future cardiovascular events for those in
the extreme quintiles were 1.62 (95% CI, 1.17-2.25) for LDL-C, 1.75 (95% CI,
1.30-2.38) for apolipoprotein A-I, 2.08 (95% CI, 1.45-2.97) for total cholesterol,
2.32 (95% CI, 1.64-3.33) for HDL-C, 2.50 (95% CI, 1.68-3.72) for apolipoprotein
B100, 2.51 (95% CI, 1.69-3.72) for non–HDL-C, and 2.98 (95%
CI, 1.90-4.67) for high-sensitivity CRP (P<.001
for trend across all quintiles). The HRs for the lipid ratios were 3.01 (95%
CI, 2.01-4.50) for apolipoprotein B100 to apolipoprotein A-I, 3.18
(95% CI, 2.12-4.75) for LDL-C to HDL-C, 3.56 (95% CI, 2.31-5.47) for apolipoprotein
B100 to HDL-C, and 3.81 (95% CI, 2.47-5.86) for the total cholesterol
to HDL-C (P<.001 for trend across all quintiles).
The correlation coefficients between high-sensitivity CRP and the lipid parameters
ranged from −0.33 to 0.15, and the clinical cut points for CRP of less
than 1, 1 to 3, and higher than 3 mg/L provided prognostic information on
risk across increasing levels of each lipid measure and lipid ratio.
Conclusions Non–HDL-C and the ratio of total cholesterol to HDL-C were as
good as or better than apolipoprotein fractions in the prediction of future
cardiovascular events. After adjustment for age, blood pressure, smoking,
diabetes, and obesity, high-sensitivity CRP added prognostic information beyond
that conveyed by all lipid measures.
While measurement of total cholesterol, low-density lipoprotein cholesterol
(LDL-C), and high-density lipoprotein cholesterol (HDL-C) are recommended
in most current cardiovascular screening algorithms,1-3 several
investigations have suggested that superior risk prediction might be achieved
by alternatively measuring apolipoproteins B100 and A-I.4-8 At
the same time, recent guidelines have emphasized the importance of non–HDL-C
as a predictor of cardiovascular risk,1 while
others have strongly advocated use of specific lipid ratios such as total
cholesterol to HDL-C, LDL-C to HDL-C, apolipoprotein B100 to apolipoprotein
A-I, and apolipoprotein B100 to HDL-C.9-12
Despite these varied recommendations, direct comparative data are sparse,
particularly among women. To address these issues, we evaluated baseline levels
of each of these lipid biomarkers as predictors of first cardiovascular event
in a large cohort of initially healthy women who were followed up prospectively
over a 10-year period. We additionally compared the predictive value of each
lipid marker to that of high-sensitivity C-reactive protein (CRP), which is
an inflammatory biomarker previously shown to predict risk in this13 and other cohorts.14-18 To
maximize the clinical utility of these data and reduce the potential for uncontrolled
confounding, we elected on an a priori basis to adjust all analyses for age,
smoking status, blood pressure, diabetes, and body mass index.
The study cohort was derived from participants in the Women’s
Health Study (WHS), a randomized, double-blind, placebo-controlled, 2 ×
2 factorial design trial of aspirin and vitamin E in the prevention of cardiovascular
disease and cancer conducted among initially healthy women aged 45 years or
were enrolled between November 1992 and July 1995 at which time they provided
baseline information on behavioral, lifestyle, and demographic risk factors.
All participants have been followed up prospectively for the occurrence of
first cardiovascular event including nonfatal myocardial infarction, nonfatal
stroke, coronary revascularization procedures, and cardiovascular-related
death. The methods of the cohort assembly, follow-up, and end-point validation
have been described previously.19-21 All
participants in the WHS provided written informed consent and the study protocol
was approved by the institutional review board of the Brigham and Women’s
Hospital (Boston, Mass).
Among WHS participants, 28 345 provided baseline blood samples
that were stored in liquid nitrogen until the time of analysis. These samples
underwent lipid analysis and evaluation for high-sensitivity CRP in a core
laboratory certified by the National Heart, Lung, and Blood Institute/Centers
for Disease Control and Prevention Lipid Standardization program. Levels of
total cholesterol and HDL-C were measured enzymatically on a Hitachi 911 autoanalyzer
(Roche Diagnostics, Basel, Switzerland) while LDL-C was determined directly
(Genzyme, Cambridge, Mass). Levels of apolipoproteins B100 and
A-I were measured by an immunoturbidometric technique on the Hitachi 911 analyzer.22 The assays for apolipoproteins A-I and B100 used standards from the World Health Organization and the International
Federation of Clinical Chemistry and Laboratory Medicine and a validation
study of these assays with those used at the Northwest Lipid Research Laboratory
(Seattle, Wash) revealed a correlation coefficient of 1.0, intercept of 0.26
mg/dL, and slope of 0.97 for apolipoprotein B100, and a correlation
coefficient of 0.99, intercept of 0.26 mg/dL, and a slope of 1.0 for apolipoprotein
A-I. High-sensitivity CRP was measured using a validated immunoturbidometric
method (Denka Seiken, Tokyo, Japan).23 Of the
samples that were received by the core laboratory, 27 748 (98%) underwent
successful evaluation for each biomarker. Non–HDL-C was calculated by
subtracting HDL-C from total cholesterol.
For the purposes of this lipid-based analysis, we followed guidelines
from the Department of Health and Human Services for lipid standardization24 and limited these analyses on an a priori basis to
women not taking hormone therapy at baseline (N = 15 632);
there were no missing laboratory data among these women. Population distributions
were computed for each biomarker and Spearman correlation coefficients were
used to discern interrelationships between the various lipid fractions and
each other, as well as high-sensitivity CRP. For each biomarker, we then divided
the baseline levels into increasing quintiles and used Cox proportional hazard
models to calculate hazard ratios (HRs) and 95% confidence intervals (CIs)
for future cardiovascular events comparing those in each of the quintiles
2 through 5 with those in the lowest (referent) quintile. Proportionality
was confirmed by Wald χ2 testing of interactions with natural
logarithm of follow-up time. Also on an a priori basis, all analyses were
simultaneously adjusted for age (years), blood pressure (Framingham categories),
diabetes, current smoking status, and body mass index (calculated as weight
in kilograms divided by the square of height in meters). Tests for trends
across quintiles of each biomarker were addressed by entering a single ordinal
term for each quintile based on the median value for that biomarker within
each quintile. The magnitude of the likelihood ratio (LR) χ2 statistic
was used to evaluate the goodness of fit of predictive models associated with
each marker individually. All analyses were controlled for randomized treatment
assignment to aspirin and vitamin E.
In a prior analysis of this cohort, we demonstrated the predictive value
of high-sensitivity CRP levels beyond that achievable by the use of LDL-C
alone.13 To extend that analysis, we sought
evidence that high-sensitivity CRP might have predictive value across all
of the lipid biomarkers measured. This latter analyses was performed after
dividing each lipid parameter into increasing tertiles and after classifying
high-sensitivity CRP levels using the clinical cut points of less than 1,
1 to 3, and higher than 3 mg/L as recommended jointly by the American Heart
Association and the Centers for Disease Control and Prevention.25 Data
analysis was conducted using SAS statistical software version 8.2 (SAS Institute
Inc, Cary, NC).
Mean (SD) age at baseline for the 15 632 initially healthy women
followed up in this study was 54.4 (7.6) years (interquartile range, 48-59
years) and the mean (SD) body mass index was 26.3 (5.3). A total of 1910 women
(12%) were current smokers, 527 (3%) had diabetes, 3847 (25%) had a history
of hypertension, and 1819 (13%) had a family history of myocardial infarction
in a parent before age 60 years.
Baseline distributions of each lipid variable as well as that of high-sensitivity
CRP are provided in Table 1. Values
are similar to those anticipated in populations of healthy middle-aged women
not taking hormone therapy.
Table 2 presents the Spearman
correlation coefficients between each lipid parameter. As expected, strong
correlations were observed between LDL-C and apolipoprotein B100 (r = 0.81), between LDL-C and non–HDL-C
(r = 0.92), between LDL-C and total cholesterol
(r = 0.91), between HDL-C and apolipoprotein
A-I (r = 0.80), between total cholesterol
and non–HDL-C (r = 0.94), and most
importantly between non–HDL-C and apolipoprotein B100 (r = 0.87). The correlation coefficients between
high-sensitivity CRP and the measured lipid parameters were r = 0.17 for total cholesterol, r = 0.15
for LDL-C, r = −0.33 for HDL-C, r = 0.27 for non–HDL-C, r = −0.19 for apolipoprotein A-I, and r = 0.29 for apolipoprotein B100.
Over an average of 10 years of follow-up, rates of completed follow-up
for the WHS exceeded 97% for morbidity and 99% for mortality. During this
period, 464 participants developed a first-ever confirmed cardiovascular end
point (131 myocardial infarction, 122 ischemic stroke, 274 coronary revascularization,
and 76 cardiovascular death, with many women having ≥2 of these end points).
To avoid double counting, only the first event for the participant was used
in these analyses.
The HRs for developing future cardiovascular events according to increasing
quintiles of each lipid variable and high-sensitivity CRP are shown in Table 3. After adjustment for age (years), blood
pressure (Framingham categories), body mass index, diabetes, and current smoking
status, each of the measured parameters was strongly associated with risk
of future cardiovascular events (P<.001 for trend
across all quintiles). The assumption of the proportional hazards model was
valid in our study in that for each biomarker, the interaction of the HR with 1n (time) was not significant (P>.05).
Of the lipid measures, the strongest association was observed with the
highly intercorrelated variables non–HDL-C and apolipoprotein B100 (Table 3). Specifically, the
fully adjusted HR for those in the highest compared with lowest baseline quintile
of non–HDL-C was 2.51 (95% CI, 1.69-3.72; LR χ2 statistic = 659.6)
whereas the comparable value for apolipoprotein B100 was 2.50 (95%
CI, 1.68-3.72; LR χ2 statistic = 660.8). Both non–HDL-C
(LR χ2 statistic = 642.2) and apolipoprotein B100 (LR χ2 statistic = 636.3) showed stronger
association than either total cholesterol or LDL-C in these data. By contrast,
the magnitude of association for HDL-C appeared to be somewhat greater than
that of apolipoprotein A-I.
High-sensitivity CRP levels also were associated with future cardiovascular
events (Table 3); the fully adjusted
HR for those in the highest compared with lowest baseline quintile of high-sensitivity
CRP was 2.98 (95% CI, 1.90-4.67; LR χ2 statistic = 650.0).
The LR χ2 statistic for high-sensitivity CRP was greater than
that of total cholesterol and LDL-C but less than that of apolipoprotein B100 and non–HDL-C, most likely reflecting modest correlations
between high-sensitivity CRP and some of the nonlipid variables used in the
multivariate adjustment. These direct comparative HRs and 95% CIs for those
in the extreme quintiles of each measured parameter are presented in Figure 1.
Comparative data on risk associated with the ratios of total cholesterol
to HDL-C, LDL-C to HDL-C, apolipoprotein B100 to apolipoprotein
A-I, and apolipoprotein B100 to HDL-C appear in Table 3. By combining information on 2 components of lipid risk
into a single clinical variable, all of these lipid ratios provided stronger
evidence of association than that achieved by any of the single variables
used alone. However, despite the apparent superiority of apolipoprotein B100 over total cholesterol and LDL-C, the lipid ratio with the strongest
association in these data was the more traditional ratio of total cholesterol
to HDL-C (top vs bottom quintile HR, 3.81; 95% CI, 2.47-5.86; P<.001 for trend; LR χ2 statistic = 694.3; Table 3). For comparison, the relative risk in
the top quintile of the ratio of apolipoprotein B100 to apolipoprotein
A-I was 3.01 (95% CI, 2.01-4.50; P<.001 for trend;
LR χ2 statistic = 674.3). These comparative data
and associated 95% CIs also are presented in Figure 1.
Cardiovascular event-free survival according to increasing quintiles
of each individual variable and ratios of total cholesterol to HDL-C and apolipoprotein
B100 to apolipoprotein A-I appear in Figure 2. The HRs and 95% CIs for future cardiovascular events after
classifying baseline high-sensitivity CRP levels by clinical cut points (<1,
1-3, and >3 mg/L) and baseline levels of non–HDL-C and apolipoprotein
B100 and ratios of total cholesterol to non–HDL-C and apolipoprotein
B100 to apolipoprotein A-I by tertile, after adjustment for age
(years), blood pressure (Framingham categories), body mass index, diabetes,
and current smoking status appear in Figure 3. Similar additive effects for high-sensitivity CRP were observed
in corresponding analyses for total cholesterol and LDL-C and the ratios of
LDL-C to HDL-C and apolipoprotein B100 to non–HDL-C (data
We observed no evidence of effect modification for high-sensitivity
CRP by age. In stratified analyses, the fully adjusted HR for those in the
highest compared with the lowest quintile of high-sensitivity CRP among women
younger than 55 years was 2.9 (P = .008)
while the corresponding HR for women aged 55 years or older was 2.8 (P = .002).
In this large, prospective cohort of initially healthy US women, we
directly compared non–HDL-C, apolipoproteins B100 and A-I,
standard lipid measures, lipid ratios, and high-sensitivity CRP as predictors
of future cardiovascular events. Overall, we observed that the magnitude of
the association was greater for apolipoprotein B100 than for either
total cholesterol or LDL-C. However, we also observed that apolipoprotein
B100 was highly correlated with non–HDL-C, and that association
for non–HDL-C was effectively equal to that of apolipoprotein B100. Moreover, the easily calculated ratio of total cholesterol to HDL-C
proved to be at least as strongly associated with cardiovascular events as
the ratio of apolipoprotein B100 to apolipoprotein A-I. In these
women, high-sensitivity CRP also was strongly associated with risk, but only
modestly correlated with any of the lipid parameters.
We believe these data have clinical relevance for several reasons. First,
with regard to individual lipid measures, our data are consistent with prior
prospective cohort studies indicating that apolipoprotein B100 is
a strong predictor of risk.4-7 However,
our data also indicate that apolipoprotein B100 is highly correlated
with non–HDL-C (r = 0.87) and that
the strength of association for non–HDL-C is clinically equivalent to
that of apolipoprotein B100. Thus, while differences between apolipoprotein
B100 and non–HDL-C may be of biological interest,7 our data do not support the use of apolipoprotein
B100 in primary risk detection because non–HDL-C can be directly
calculated by subtracting HDL-C from total cholesterol at no incremental cost
beyond usual lipid evaluation. In this regard, our data for apolipoprotein
B100 are consistent with findings from the Atherosclerosis Risk
in Communities (ARIC) study in which apolipoprotein B100 was a
univariate predictor of risk but did not substantively contribute to risk
beyond that achievable with standard lipid fractions.26 Consistent
with both the ARIC investigation and the Quebec Cardiovascular Study,5 we observed no substantive incremental clinical benefit
of apolipoprotein A-I evaluation over that of HDL-C.
Second, with regard to lipid ratios, our data are consistent with prior
reports that the ratio of apolipoprotein B100 to apolipoprotein
A-I is strongly associated with incident cardiovascular events independent
of the nonlipid covariates typically used in global risk prediction scores.4-8 However,
in a manner parallel to that observed for apolipoprotein B100 alone,
we also observed that the strength of association for the ratio of apolipoprotein
B100 to apolipoprotein A-I was not superior in these data to the
ratio of total cholesterol to HDL-C. Thus, on the basis of the data in this
large prospective cohort of initially healthy women as well as other nested
case-control studies that have found the ratio of total cholesterol to HDL-C
to perform favorably,12,27 it
would not seem clinically important to replace standard lipid measures with
more complex apolipoprotein evaluations—at least for the purpose of
primary risk detection. On the other hand, our data and those of several prior
studies do suggest that the use of either the ratio of total cholesterol to
HDL-C or LDL-C to HDL-C is superior to the use of total cholesterol or LDL-C
alone, and thus do not support the position taken by the European SCORE project,
which advocates the use of total cholesterol in isolation.3
In the current data, the HR for those in the top vs bottom quintile
of high-sensitivity CRP was 2.98 (95% CI, 1.90-4.67), which was greater than
that of both non–HDL-C (HR, 2.51; 95% CI, 1.69-3.72) and apolipoprotein
B100 (HR, 2.50; 95% CI, 1.68-3.72). In contrast to the strong correlations
observed between lipid measures, the correlation coefficients between high-sensitivity
CRP and each lipid parameter were smaller and ranged from −0.33 to 0.15.
These observations are consistent with the hypothesis that both inflammation
and hyperlipidemia contribute jointly to the atherothrombotic process. At
the same time, the LR statistic in these data for high-sensitivity CRP was
greater than that of total cholesterol and LDL-C, but less than that of non–HDL-C
and apolipoprotein B100. This latter observation suggests somewhat
greater correlations for high-sensitivity CRP than for standard lipids with
the nonlipid variables used in our multivariable adjustments, which included
both diabetes and obesity. These effects are not surprising because high-sensitivity
CRP levels also predict the onset of type 2 diabetes and adipocytes are a
Our analysis used state-of-the art assays for all measures (including
apolipoproteins A-I and B100) and thus are unlikely to be affected
by laboratory issues that have been raised in some earlier studies of both
lipid and inflammatory biomarkers. Furthermore, the large-scale prospective
cohort approach taken greatly reduces the possibility of chance and bias as
alternative explanations for our findings.
However, our study does have some limitations that merit consideration.
We evaluated plasma levels only once and thus our data may be susceptible
to intra-individual variation. Our study also evaluated middle-aged women
and it is known that adverse effects on several blood variables can occur
during and after menopause. However, our main findings are consistent with
those from a nested case-control study performed within the Nurses’
Health Study in which the average age of women evaluated was older.27 Our lipid-based data are also consistent with prior
studies conducted predominantly or exclusively among men.12,26 Finally,
we do not have full assessment of triglyceride levels among these women and
thus cannot evaluate whether any of the other parameters measured might have
greater or lesser value among those with altered triglyceride patterns.
While our analyses support the use of standard lipid measures rather
than apolipoproteins A-I and B100 in primary risk detection, these
data should not be construed to exclude a potential role for apolipoprotein
B100 or the ratio of apolipoprotein B100 to apolipoprotein
A-I in monitoring patients taking statins. In this regard, there was virtually
no use of statin therapy at the time of enrollment into the WHS and overall
usage rates remained very low at all points of follow-up reflecting the low-risk
nature of this cohort.19 However, in the Air
Force Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS), a
randomized trial of lovastatin, the levels of apolipoprotein B100 when
participants were receiving treatment and the ratio of apolipoprotein B100 to apolipoprotein A-I were better predictors of future cardiovascular
events than LDL-C.30 Levels of apolipoproteins
B100 and A-I while participants were receiving treatment also have
been found to significantly predict recurrent cardiovascular events in several
secondary prevention studies.8,31,32 It
has been hypothesized on this basis that the monitoring of apolipoprotein
B100 could replace the current standard lipid profile evaluation
among patients taking statins.33 With regard
to high-sensitivity CRP, recent data suggest potential utility for this inflammatory
biomarker as an adjunctive method to monitor statin efficacy not as a replacement
for LDL-C.34-36 The
possible role of combined apolipoprotein and high-sensitivity CRP evaluation
to monitor patients taking statins needs to be evaluated.
Corresponding Author: Paul M Ridker, MD,
Center for Cardiovascular Disease Prevention, Brigham and Women’s Hospital,
900 Commonwealth Ave E, Boston, MA 02215 (firstname.lastname@example.org).
Author Contributions: Drs Ridker, Buring, and
Cook had full access to all of the data in the study and take responsibility
for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Ridker, Rifai.
Acquisition of data: Ridker, Rifai, Bradwin,
Analysis and interpretation of data: Ridker,
Drafting of the manuscript: Ridker.
Critical revision of the manuscript for important
intellectual content: Rifai, Cook, Bradwin, Buring.
Statistical analysis: Cook.
Obtained funding: Ridker, Buring.
Administrative, technical, or material support:
Study supervision: Ridker, Rifai.
Financial Disclosures: Dr Ridker is listed
as a coinventor on patents held by the Brigham and Women’s Hospital
that relate to the use of inflammatory biomarkers in cardiovascular disease.
No other authors reported financial disclosures.
Funding/Support: The research for this article
was supported by grants from the Donald W. Reynolds Foundation (Las Vegas,
NV), the Leducq Foundation (Paris, France), and the Doris Duke Charitable
Foundation (New York, NY). The Women’s Health Study is supported by
grants from the National Heart, Lung, and Blood Institute and the National
Role of the Sponsor: The funding agencies played
no role in the design, conduct, data management, analysis, or manuscript preparation
related to this project.
Create a personal account or sign in to: