Context Several novel risk factors for atherosclerosis have recently been proposed,
but few comparative data exist to guide clinical use of these emerging biomarkers.
Objective To compare the predictive value of 11 lipid and nonlipid biomarkers
as risk factors for development of symptomatic peripheral arterial disease
(PAD).
Design, Setting, and Participants Nested case-control study using plasma samples collected at baseline
from a prospective cohort of 14 916 initially healthy US male physicians
aged 40 to 84 years, of whom 140 subsequently developed symptomatic PAD (cases);
140 age- and smoking status–matched men who remained free of vascular
disease during an average 9-year follow-up period were randomly selected as
controls.
Main Outcome Measure Incident PAD, as determined by baseline total cholesterol, high-density
lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C),
total cholesterol–HDL-C ratio, triglycerides, homocysteine, C-reactive
protein (CRP), lipoprotein(a), fibrinogen, and apolipoproteins (apo) A-I and
B-100.
Results In univariate analyses, plasma levels of total cholesterol (P<.001), LDL-C (P = .001), triglycerides
(P = .001), apo B-100 (P
= .001), fibrinogen (P = .02), CRP (P = .006), and the total cholesterol–HDL-C ratio (P<.001) were all significantly higher at baseline among men who
subsequently developed PAD compared with those who did not, while levels of
HDL-C (P = .009) and apo A-I (P = .05) were lower. Nonsignificant baseline elevations of lipoprotein(a)
(P = .40) and homocysteine (P
= .90) were observed. In multivariable analyses, the total cholesterol–HDL-C
ratio was the strongest lipid predictor of risk (relative risk [RR] for those
in the highest vs lowest quartile, 3.9; 95% confidence interval [CI], 1.7-8.6),
while CRP was the strongest nonlipid predictor (RR for the highest vs lowest
quartile, 2.8; 95% CI, 1.3-5.9). In assessing joint effects, addition of CRP
to standard lipid screening significantly improved risk prediction models
based on lipid screening alone (P<.001).
Conclusions Of 11 atherothrombotic biomarkers assessed at baseline, the total cholesterol–HDL-C
ratio and CRP were the strongest independent predictors of development of
peripheral arterial disease. C-reactive protein provided additive prognostic
information over standard lipid measures.
Several novel risk factors have been proposed as potential criteria
for improved detection of subclinical atherosclerosis. In particular, clinical
interest has focused on emerging lipid parameters such as lipoprotein(a),
apolipoprotein (apo) A-I, and apo B-100; on inflammatory biomarkers such as
C-reactive protein (CRP) and fibrinogen; and on nutritional markers associated
with premature atherothrombosis, such as total plasma homocysteine. Direct
comparisons of the magnitude of predictive value associated with each of these
parameters have been rare. Furthermore, it has been uncertain whether screening
for any of these novel biomarkers substantially improves risk prediction over
that associated with standard lipid screening alone.1-3
Recently, we compared a series of lipid and nonlipid risk factors for
atherosclerosis in a prospective cohort of apparently healthy middle-aged
women.4 In that study, we found that the addition
of CRP to screening based on standard lipid levels improved identification
of women at increased risk for future myocardial infarction or stroke. In
contrast, few of the other biomarkers evaluated significantly improved risk
prediction models.
Due to clinical interest in these data and their implications for population-based
screening programs, we elected to directly compare the magnitude of predictive
value associated with a large series of lipid and nonlipid risk factors in
a prospective cohort of apparently healthy men. We elected to use as our clinical
end point the future development of symptomatic peripheral arterial disease
(PAD), a disorder for which few data exist regarding the potential usefulness
of novel risk factors for atherosclerosis.
The study population consisted of apparently healthy middle-aged men
participating in the Physicians' Health Study, a prospective, randomized trial
of aspirin and beta-carotene in the primary prevention of cardiovascular disease
and cancer.5 In brief, 14 916 men aged
40 to 84 years who had no prior history of cardiovascular disease or cancer
provided baseline plasma samples that were collected in EDTA and then frozen
at –80°C until the time of analysis. Participants were monitored
over an average follow-up period of 9 years for the occurrence of incident
health events, including the development of intermittent claudication and
hospitalizations for peripheral arterial revascularization procedures. For
the purposes of this analysis, case subjects were defined as those apparently
healthy men who subsequently developed either of these PAD end points during
the study period; there were 140 such subjects. By contrast, 140 control subjects
were selected at random from the remaining study subjects who remained free
of reported cardiovascular disease. Controls were matched to cases on the
basis of age, smoking status, and length of follow-up. No study participant
had a baseline history of intermittent claudication or prior lower extremity
revascularization procedures. Because the study design focused on symptomatic
lower extremity PAD, participants who underwent revascularization of either
the renal or carotid arteries were not considered as case subjects.
Measurement of Risk Factors
Baseline plasma samples from case and control subjects were thawed and
assayed for each plasma parameter using commercially available analytic systems.
Total cholesterol (TC), high-density lipoprotein (HDL) cholesterol, direct
low-density lipoprotein (LDL) cholesterol, and triglycerides were assayed
on a Hitachi 911 analyzer (Roche Diagnostics, Indianapolis, Ind) using reagents
from Roche Diagnostics and Genzyme Corporation (Cambridge, Mass). Total plasma
homocysteine was measured with an IMx homocysteine assay on an IMx analyzer
(Abbott Laboratories, Abbott Park, Ill). High-sensitivity CRP and lipoprotein(a)
were assayed using a latex-enhanced immunonephelometric assay on a BN II analyzer
(Dade Behring, Newark, Del),6 while fibrinogen,
apo A-I, and apo B-100 were simultaneously measured on this device by immunoassay.
To ascertain the potential clinical usefulness of each putative risk
marker, we followed an a priori analysis plan in which the distribution for
each parameter was first compared between case and control groups using the
nonparametric Wilcoxon rank sum test. After dividing subjects into quartiles
based on the distribution of control values, we then used conditional logistic
regression analyses to compute relative risks (RRs) of future PAD associated
with increasing levels of each study parameter. By so doing, the magnitude
of risk associated with levels in the highest vs the lowest quartile could
be directly compared for each of the 11 parameters. Multivariable estimates
of risk were computed in similar models which, in addition to the matching
variables of age and smoking, adjusted further for hypertension, body mass
index, family history of premature atherosclerosis, diabetes, and exercise
frequency. Finally, we addressed the potential additive value of each novel
parameter to standard lipid screening in a 2-step process. Initially, for
any marker found to be positively associated with risk of PAD in univariate
analyses, we used likelihood ratio testing to determine whether the addition
of that marker to standard cholesterol screening significantly improved risk
prediction models. Similar logistic regression analyses were then performed
after dividing study subjects into 4 groups based on median levels for standard
lipid markers and on median levels for each novel risk parameter. All P values are 2-tailed and all confidence intervals (CIs)
computed at the 95% level (SAS Version 8.2, SAS, Cary, NC).
Due to matching, mean age and smoking patterns were virtually identical
in the case and control groups. As expected, traditional atherosclerotic risk
factors such as diabetes, hypertension, and a family history of cardiovascular
disease were more prevalent at baseline among those who subsequently developed
symptomatic PAD (cases) compared with those who did not (controls) (Table 1).
The mean time to diagnosis of incident PAD for the study population
was 60 months (range, 3-155 months). As shown in Table 2, median levels of TC, LDL cholesterol, triglycerides, apo
B-100, and the TC/HDL-C ratio were all significantly higher at baseline among
case subjects. In contrast, HDL-C levels were significantly lower among those
who subsequently developed PAD while levels of apo A-I were marginally reduced
and no significant difference was observed for lipoprotein(a). Median levels
of the 2 inflammatory markers, CRP and fibrinogen, were significantly higher
at baseline among case subjects compared with controls. No significant difference
was observed for total plasma homocysteine.
Age- and smoking-matched RRs of developing PAD for increasing quartiles
of each of the measured parameters are presented in Table 3. Among the lipid parameters, TC, HDL-C, LDL-C, triglycerides,
and apo B-100 were all significant predictors of risk, although the TC/HDL-C
ratio was the single strongest predictor (highest quartile vs lowest quartile:
RR, 3.4; 95% CI, 1.7-7.0). A modest but nonsignificant trend was observed
for apo A-I, while there was no discernable risk gradient in these data for
lipoprotein(a).
Among the nonlipid parameters, CRP was the single strongest predictor
(highest quartile vs lowest quartile: RR, 2.5; 95% CI, 1.3-5.0) (Table 3). Also, increasing quartiles of
fibrinogen were a significant predictor of risk of developing PAD. There was
no discernable risk gradient for homocysteine.
None of these associations were materially altered in analyses which,
in addition to the matching variables of age and smoking status, additionally
controlled for hypertension, body mass index, family history of premature
atherosclerosis, exercise frequency, and diabetes (Table 4). In these adjusted analyses, the strongest lipid predictor
remained the TC/HDL-C ratio (highest vs lowest quartile: RR, 3.9; 95% CI,
1.7-8.6) while the strongest nonlipid predictor remained CRP (highest vs lowest
quartile: RR, 2.8; 95% CI, 1.3-5.9).
Similar findings, although with wider CIs, were observed in analyses
limited to participants who underwent peripheral arterial revascularization
procedures in addition to developing intermittent claudication. For example,
in this subgroup of 30 participants, the RR of incident PAD for those with
baseline TC/HDL-C ratios in the highest vs the lowest quartile was 3.2 (95%
CI, 1.0-10.9) while that for CRP was 2.1 (95% CI, 0.8-5.6). In these subgroup
analyses, as in the study as a whole, statistically significant trends in
risk were observed across quartiles of TC (P = .01),
CRP (P = .02), apo B-100 (P
= .02), LDL-C (P = .02), and triglycerides (P = .03).
To explore whether any of the novel risk parameters added to the predictive
value of standard lipid-based screening, a series of additional analyses were
undertaken. First, likelihood ratio tests were used to compare the fit of
risk prediction models based on the measurement of each novel marker in combination
with standard lipid screening. Second, we computed the RR of developing PAD
in analyses that stratified study participants into 4 groups based on median
lipid levels and on median levels for each of the other measured parameters
(Figure 1). To address the robustness
of these findings, analyses were performed separately using either TC or the
TC/HDL-C ratio as the method for standard lipid screening.
As might be expected due to intercorrelation among many of the lipid
parameters, there was little evidence of benefit of adding LDL-C, apo A-I,
or apo B-100 to analyses that already incorporated the TC/HDL-C ratio. Similarly,
there was minimal evidence of benefit of adding either lipoprotein(a) or homocysteine
to standard lipid screening with either TC or the TC/HDL-C ratio.
In contrast, both of the inflammatory markers (CRP and fibrinogen) significantly
improved the predictive value of models based on either TC (P<.001 for CRP; P = .01 for fibrinogen)
or the TC/HDL-C ratio (P<.001 for CRP; P = .04 for fibrinogen). These effects were stronger for CRP than for
fibrinogen, although the magnitude of this difference was small.
In a post hoc analysis evaluating the addition of both CRP and fibrinogen
to standard lipid screening, there was little evidence of any further improvement
in risk prediction models, most likely because levels of CRP and fibrinogen
were also intercorrelated (r = 0.6; P = .001).
In this prospective evaluation of 11 plasma biomarkers associated with
the development of PAD, the TC/HDL-C ratio proved to be the single strongest
lipid predictor of risk. Indeed, once this ratio was taken into account we
found little evidence that additional screening for other lipid parameters
including LDL-C, apo A-I, apo B-100, or lipoprotein(a) had significant clinical
usefulness. However, the addition of either CRP or fibrinogen to standard
lipid screening significantly improved the predictive value of computed risk
prediction models. Of these 2 intercorrelated inflammatory variables, CRP
was the stronger univariate predictor of risk and had the greater additive
value when combined with either TC or the TC/HDL-C ratio, although the magnitude
of these latter differences was small. Finally, we found little efficacy for
homocysteine evaluation in these data, either alone or in combination with
standard lipid screening.
The current data for CRP were generated with a commercially available
high-sensitivity assay6 and thus corroborate
prior results from these study participants in which CRP was measured using
an in-house research-based assay.7 Our observation
that CRP testing combined with standard lipid screening appears to provide
an improved method of detecting subclinical atherosclerosis also confirms
prior work for myocardial infarction in men8-10
and for coronary heart disease and stroke in women.4
To our knowledge, these data are the first to directly address the magnitude
of predictive value of novel risk factors associated with the development
of symptomatic PAD. We believe the strongly positive findings in this study—particularly
those for the TC/HDL-C ratio, CRP, and fibrinogen—merit careful consideration
among clinicians interested in programs of general population-based screening
to improve the detection of subclinical atherosclerosis. However, we also
believe that our null data for homocysteine and lipoprotein(a) should not
be construed as suggesting no role for these latter markers in the detection
of vascular risk. Recent overview analyses for both homocysteine11
and lipoprotein(a)12 have reported statistically
significant relationships, albeit with a magnitude of effect smaller than
that observed for the inflammatory markers CRP and fibrinogen. Thus, despite
the null data presented here, clinical evaluation for homocysteine and lipoprotein(a)
might nonetheless be considered in the setting of markedly premature atherosclerosis
or when there is a strong family history of atherothrombosis in the absence
of other major risk factors. Homocysteine evaluation is also likely to have
greater yield among patients with renal failure.11
On the other hand, our null data for homocysteine and lipoprotein(a) argue
against population-based screening for these parameters, and as such are in
accordance with recent recommendations from the American College of Cardiology
and the American Heart Association.2,3,11
The use of self-reported symptomatic PAD as our primary a priori end
point represents a potential limitation of these data. However, we believe
this clinically relevant end point is valid for several reasons. First, our
study participants were physicians, a group in whom validation rates for several
other self-reported vascular and nonvascular end points have consistently
been excellent.13-16
In particular, among Physicians' Health Study participants, self-reported
incident angina pectoris has been validated by either exercise stress testing
or positive angiography in more than 98% of cases.13
Second, any potential misclassification on this basis would tend, if anything,
to bias these data toward the null, not a false-positive finding. Finally,
in the subgroup of study participants who not only self-reported incident
intermittent claudication but who also underwent lower extremity revascularization,
the point estimates of effect were minimally altered. Thus, we believe the
use of self-reported symptomatic disease in our study is not only valid, but
represents a true clinical end point rather than a surrogate marker for disease.
The current study was specifically designed to assess the magnitude
of predictive value of several inexpensive plasma biomarkers for systemic
atherosclerosis as well as their potential usefulness as adjuncts to standard
lipid screening. It is important to note, however, that bedside techniques
that assess peripheral blood flow also have clinical usefulness in the detection
of PAD. In particular, simple physical examination of the distal pulses as
well as assessment of the ankle-brachial index are useful for the detection
of clinically relevant peripheral arterial obstruction and for the prediction
of subsequent vascular events.17,18
Other modalities such as magnetic resonance imaging or electron beam computed
tomography have also been advocated as tools for the detection of subclinical
atherosclerosis. However, in marked contrast to physical examination or the
measurement of a small panel of plasma-based biomarkers, these direct imaging
techniques are quite expensive, a major concern for primary prevention screening
programs. Thus, carefully designed studies comparing plasma biomarkers with
direct imaging techniques will be needed to assess the clinical efficacy and
cost-effectiveness of these different approaches.1,19
1.Ridker PM. Evaluating novel cardiovascular risk factors.
Ann Intern Med.1999;130:933-937.Google Scholar 2.Smith SC, Greenland P, Grundy S. Beyond secondary prevention: identifying the high-risk patient for
primary prevention: executive summary.
Circulation.2000;101:111-116.Google Scholar 3.Grundy SM, Bazzarre T, Cleeman J.
et al. Beyond secondary prevention: identifying the high-risk patient for
primary prevention: medical office assessment.
Circulation.2000;101:e3-e11.Google Scholar 4.Ridker PM, Hennekens CH, Buring JE, Rifai N. C-reactive protein and other markers of inflammation in the prediction
of cardiovascular disease in women.
N Engl J Med.2000;342:836-843.Google Scholar 5.Steering Committee of the Physicians' Health Study Research Group. Final report of the aspirin component of the ongoing Physicians' Health
Study.
N Engl J Med.1989;321:129-135.Google Scholar 6.Rifai N, Tracy RP, Ridker PM. Clinical efficacy of an automated high-sensitivity C-reactive protein
assay.
Clin Chem.1999;45:2136-2141.Google Scholar 7.Ridker PM, Cushman M, Stampfer MJ, Tracy RP, Hennekens CH. Plasma concentration of C-reactive protein and risk of developing peripheral
vascular disease.
Circulation.1998;97:425-428.Google Scholar 8.Ridker PM, Cushman M, Stampfer MJ, Tracy RP, Hennekens CH. Inflammation, aspirin, and the risk of cardiovascular disease in apparently
healthy men.
N Engl J Med.1997;336:973-979.Google Scholar 9.Koenig W, Sund M, Frohlich M.
et al. C-reactive protein, a sensitive marker of inflammation, predicts future
risk of coronary heart disease in initially healthy middle-aged men.
Circulation.1999;99:237-242.Google Scholar 10.Ridker PM, Glynn RJ, Hennekens CH. C-reactive protein adds to the predictive value of total and HDL cholesterol
in determining risk of first myocardial infarction.
Circulation.1998;97:2007-2011.Google Scholar 11.Malinow MR, Bostom AG, Krauss RM. Homocyst(e)ine, diet, and cardiovascular disease.
Circulation.1999;99:178-182.Google Scholar 12.Danesh J, Collins R, Peto R. Lipoprotein(a) and coronary heart disease.
Circulation.2000;102:1082-1085.Google Scholar 13.Carmargo CA, Stampfer MJ, Glynn RJ.
et al. Moderate alcohol consumption and risk for angina pectoris or myocardial
infarction in US male physicians.
Ann Intern Med.1997;126:372-375.Google Scholar 14.Ascherio A, Rimm EB, Giovannucci EL.
et al. A prospective study of nutritional factors and hypertension among US
men.
Circulation.1992;86:1475-1484.Google Scholar 15.Christen WG, Glynn RJ, Seddon JM, Manson JE, Buring JE, Hennekens CH. Confirmation of self-reported cataract in the Physicians' Health Study.
Ophthalmic Epidemiol.1994;1:85-91.Google Scholar 16.Klag MJ, He J, Mead LA, Ford DE, Pearson TA, Levine DM. Validity of physicians' self-reports of cardiovascular disease risk
factors.
Ann Epidemiol.1993;3:442-447.Google Scholar 17.Criqui MH, Langer RD, Fronek A.
et al. Mortality over a period of 10 years in patients with peripheral arterial
disease.
N Engl J Med.1992;326:381-386.Google Scholar 18.Vogt MT, Cauley JA, Newman AB.
et al. Decreased ankle/arm blood pressure index and mortality in elderly women.
JAMA.1993;270:465-469.Google Scholar 19.Ridker PM. High-sensitivity C-reactive protein: potential adjunct for global risk
assessment in the primary prevention of cardiovascular disease.
Circulation.2001;103:1813-1818.Google Scholar