Clinical Chemistry 58:1

72–82 (2012) Reviews

Biomarkers and Cardiovascular Risk Assessment for

Primary Prevention:
An Update
Lauren G. Gilstrap1 and Thomas J. Wang1*

BACKGROUND: Interest in cardiovascular biomarkers in and molecular research, which have yielded insight
primary prevention has increased dramatically in the about early cardiovascular pathophysiology and simul-
past decade. This increase has been fueled by an im- taneously provided novel platforms for biomarker dis-
proved understanding of cardiovascular pathophysiol- covery (1 ). Another factor contributing to the in-
ogy, as well as novel technologies for biomarker creased focus on early screening tests has been the
identification. recognition that traditional cardiovascular risk factors
(e.g., hypertension, hyperlipidemia, smoking, diabe-
CONTENT: In this review we provide a brief overview of tes) do not fully explain interindividual variation in
recent concepts in the evaluation of screening bio- cardiovascular risk. For instance, a large proportion of
markers, because biomarkers may behave differently individuals who develop cardiovascular disease have
when used for screening as opposed to diagnosis or few or no risk factors (2 ).
disease staging. The following specific biomarker ex- The challenge for physicians, researchers, and
amples are then discussed, with a focus on data from clinical chemists is to develop screening strategies that
primary prevention studies: high-sensitivity C-reactive safely, accurately, and cost-effectively identify individ-
protein, B-type natriuretic peptide, lipoprotein- uals at risk for cardiovascular events well before symp-
associated phospholipase A2, and high-sensitivity tro- toms appear. Preventive interventions are likely to be
ponin T. The article concludes by addressing novel most effective in this period, because atherosclerosis
platforms for biomarker discovery, reviewing recent can take decades to manifest. The measurement of
examples from the field of metabolomics. “novel” biomarkers could be an important component
of these strategies. Although a biomarker can be any-
SUMMARY: An ongoing challenge is to develop screen- thing that reflects a biological process—from genetic
ing strategies that can identify individuals at risk for markers to imaging tests—soluble biomarkers are par-
cardiovascular events well before symptoms appear. ticularly attractive because they are easy to obtain and
For this purpose, the measurement of soluble biomark- generally reproducible (3 ).
ers could be an important adjunct to traditional car- Because biomarkers may behave differently when
diovascular risk assessment. Recent studies highlight used for screening as opposed to diagnosis or disease
both the strengths and limitations of “novel” circulat- staging, a distinct set of metrics has arisen to assess
ing biomarkers, and suggest that substantial work is screening biomarkers. Thus, we begin this review with
still needed to identify biomarkers that are sufficiently a brief overview of recent concepts in the evaluation of
accurate and cost-effective for routine use in primary screening biomarkers. Then, we describe a few specific
prevention. examples from the recent literature. For the sake of this
© 2011 American Association for Clinical Chemistry review, only circulating biomarkers will be discussed.

General Considerations
Interest in cardiovascular biomarkers in primary pre-
vention has increased dramatically in the past decade. In 2007, Morrow and de Lemos outlined 3 criteria for
This increase has been fueled by advances in genetic evaluating novel biomarkers: (a) ease of measurement;
(b) addition of information; and (c) effect on manage-
ment (4 ). The first criterion is typically straightforward
1
to assess. Indeed, one of the advantages of soluble bio-
Cardiology Division and Department of Medicine, Massachusetts General Hos-
pital, Harvard Medical School, Boston, MA.
markers is that standardized, reproducible assays exist
* Address correspondence to this author at: Cardiology Division, GRB-800, Mas- for most biomarkers of interest. In contrast, addressing
sachusetts General Hospital, Boston, MA 02114. Fax 617-726-4105; e-mail the second criterion—the incremental information
tjwang@partners.org.
Received August 1, 2011; accepted September 16, 2011. from measuring a biomarker— can be challenging. In
Previously published online at DOI: 10.1373/clinchem.2011.165712 cardiology, a biomarker is useful only when it contrib-

72
Biomarkers in Primary Prevention
Reviews

utes information in addition to that provided by tradi- Discrimination is the ability of a biomarker to dis-
tional risk factors. The best way to evaluate this contri- tinguish those who develop a disease from those who
bution statistically in low-risk populations has been a do not (16 ). Discrimination can be characterized by
source of debate; several approaches have been pro- the area under the ROC curve (AUC) or the c-statistic.
posed, which are described in the following section. The c-statistic is a function of the sensitivity and spec-
Whether a biomarker is used to screen for a dis- ificity of a test across all diagnostic thresholds. The sen-
ease, diagnose a disease, or inform prognosis, it should sitivity of a test refers to its ability to detect disease
affect clinical management. Because few randomized when it is present, the so-called true-positive rate.
trials have been performed to evaluate biomarkers in Specificity refers to the ability of a test to exclude dis-
the primary prevention setting, little is known about ease when it is not present, the true-negative rate. The
the impact of biomarker-guided strategies on clinical c-statistic ranges from 0.5 (uninformative test) to 1.0
outcomes. The section on high-sensitivity C-reactive (perfect discrimination) (17 ). In general, a c-statistic
protein (CRP)2 highlights some of the challenges in- ⬎0.7 is considered good. The c-statistic for coronary
volved in producing the necessary evidence. heart disease (CHD) with traditional risk factors (e.g.,
A specific consideration in measuring biomarkers in the form of the Framingham Risk Score) is approx-
for cardiovascular screening is the potential impor- imately 0.75 (18 ). Thus, the value of a new biomarker
tance of noncardiac sources of variation, given the need can be gauged by determining how much higher the
to interpret concentration differences that are typically c-statistic becomes with the combination of traditional
much smaller than those observed in acutely ill pa- risk factors and the biomarker test. A test that increases
tients. Such variation may be particularly relevant for the c-statistic by 0.05 or more is thought to add clini-
biomarkers originating in extracardiac tissue, such as cally useful information (19 ). There is no consensus
CRP, which increases with infectious states or weight about the importance of lesser changes in the c-statistic
gain. Nonetheless, it can also be important for cardiac- (0.01– 0.05), which may depend in part on the baseline
derived biomarkers such as B-type natriuretic peptide value. For instance, moving from 0.75 (with the Fra-
(BNP), which is influenced by noncardiac factors such mingham Risk Score) to 0.77 (with the Framingham
as sex, (5, 6 ) sex hormone status, (7, 8 ) and weight Risk Score plus biomarkers) may be more notable than
(9 –13 ). An understanding of such effects is essential moving from 0.50 to 0.55.
for implementing biomarker measurements in the am- Calibration refers to the concordance between
bulatory setting. predicted risk and observed risk, assessed by compar-
ing modeled risk estimates with actual event rates (20 ).
Statistical Assessment The most commonly reported calibration statistic is
the Hosmer-Lemeshow statistic. Tests or risk models
A statistically significant association between a bio- with a Hosmer-Lemeshow P value ⬎0.05 are consid-
marker and cardiovascular disease is necessary but
ered well calibrated because there is no significant dif-
not sufficient to determine if the biomarker is clini-
ference between predicted and observed event rates.
cally relevant (14 ). Statistical significance, in this
Predicted risk is clinically relevant, because treatment
case, indicates only that the mean biomarker con-
decisions are often based on estimation of future risk
centration differs between individuals with and
(20 ). If a patient’s risk of having a cardiovascular event
without the disease (or between those who have and
is quite high, a clinician may be more likely to start
don’t have an event). Because the distributions of
early pharmacologic therapy to lower LDL (with
actual biomarker concentrations between diseased
statins) or reduce thrombotic tendency (with aspirin).
and nondiseased groups may overlap substantially,
On the other hand, if the patient’s risk is low, the side
the value of a biomarker measurement in a given
effects might outweigh the likely benefits of pharmaco-
individual may be limited, even if a significant asso-
ciation exists in the overall sample (15 ). Therefore, at logic therapy.
least 3 additional statistical criteria have been proposed Estimates of calibration are sensitive to differing
to assess the utility of new biomarkers: discrimination, baseline levels of risk. For instance, if a given risk model
calibration, and reclassification (Table 1). is derived in a high-risk population but tested in a low-
risk population, the predicted risk estimates might be
unreliable. Recalibration of the risk model by adjusting
the baseline risk estimates to fit the target population
may help to mitigate over- or underestimation of risk
2
Nonstandard abbreviations: CRP, C-reactive protein; BNP, B-type natriuretic (21 ). Unfortunately, implementing this method in the
peptide; AUC, area under the ROC curve; NRI, net reclassification improvement;
CHD, coronary heart disease; Lp-PLA2, lipoprotein-associated phospholipase a2; clinical setting may not be feasible. Providers typically
hs, high sensitivity; NT, N-terminal; MS, mass spectrometry. have access only to published risk scores and do not

Clinical Chemistry 58:1 (2012) 73

Reviews

Table 1. Selected criteria for a ssessing the incremental predictive value of new biomarkers.a

Metric(s) Attribute Description Advantages Limitations

c-Statistic, AUC Discrimination Captures the sensitivity and • Easily understood • Difficult for new
specificity of a marker, biomarkers to raise the
across a full range of c-statistic when
cutpoints existing variables
discriminate well
• No reliance on specific
thresholds or categories
Hosmer-Lemeshow Calibration Assesses concordance • Assesses accuracy of predicted • Relatively insensitive to
statistic between predicted and risks (basis for clinical differences between
observed event rates in decisions) models
longitudinal data
• Provides a global
measure, even though
clinical decisions may
be based on predicted
risks within a relatively
narrow range
Net reclassification Reclassification Assesses the proportion of • Clinically relevant when risk • Sensitive to changes in
improvement individuals reclassified categories are linked to the number of risk
correctly by the addition treatment decisions categories and choice
of new biomarker(s) of cut points
• Incorporates the accuracy of • Gives same weight to
reclassification reclassifications that
are unlikely to affect
clinical decisions
a
Adapted from Wang et al. (16 ), with permission.

have the luxury of recalibrating risk scores to reflect zation of a given patient, it could affect their
their individual patient populations (16 ). management.
Reclassification refers to the ability of a test to Reclassification can be described by estimating the
change an individual’s risk classification. Because re- proportion of individuals in a population who are re-
classification is based on risk categories, it is potentially classified, although this metric doesn’t account for
the most clinically relevant criterion, because treat- whether the reclassification is “correct” or not. Pencina
ment decisions are often linked to whether a patient is and colleagues have proposed a metric known as the
considered high risk or low risk rather than on their net reclassification improvement (NRI), which sum-
specific predicted event rate. On the other hand, the marizes the net proportion of individuals with “cor-
usefulness of reclassification relies on the existence of rect” reclassification (e.g., those who develop events
well-established risk categories. For cardiovascular dis- who were up-classified, and those who don’t develop
ease, the Adult Treatment Panel III is the most com- events who were down-classified) and “incorrect” re-
monly used risk stratification algorithm (22 ). It incor- classification (those who develop events who were
porates the traditional risk factors of the Framingham down-classified, and those who don’t develop events
Risk Score and categorizes individuals as low, interme- who were up-classified) (24 ). They have recently ex-
diate, or high risk, on the basis of their 10-year pre- tended this concept to include a “category-free” NRI,
dicted risk of CHD. Low-risk individuals are those who which does not depend on the existence of fixed risk
have a ⬍10% predicted risk of a CHD event over 10 categories (25 ).
years. Intermediate- risk individuals have a 10%–19% There are advantages and limitations to each cri-
predicted risk, and high-risk individuals have a 20% or terion for evaluating biomarkers. The c-statistic (dis-
greater predicted risk (23 ). According to the Adult crimination) is easily understood, in that it is a func-
Treatment Panel III, individuals in the high-risk group tion of sensitivity and specificity and does not rely on
warrant more aggressive medical intervention, specifi- specific thresholds or categories. Nonetheless, it is of-
cally initiation of a statin with an LDL goal of ⬍70 ten quite challenging for a new biomarker to raise the
mg/dL (14 ). If the addition of one or more biomarkers c-statistic when existing models, such as the Framing-
to traditional risk factors changed the risk categori- ham Risk Score, already discriminate well. Biomarkers

74 Clinical Chemistry 58:1 (2012)

Biomarkers in Primary Prevention
Reviews

could provide clinically important information despite groups. For instance, Macy et al. found an intraindi-
small increments in the c-statistic. The Hosmer- vidual variability (CVI) of 42%, compared with a
Lemeshow statistic (calibration) assesses the accuracy between-individual variability (CVG) of 93% (31 ). The
of predicted risk estimates, which are the basis for most index of individuality (CVI/CVG) of 0.46 is comparable
clinical decisions. However, as noted, recalibration of to other analytes such as total cholesterol. Another
models may be necessary when applied to different study, involving Framingham Heart Study partici-
populations. Furthermore, as with discrimination, it pants, examined the rank stability of long-term CRP
may be difficult to measure improvement in calibra- measurements. Between 2 routine examinations (a
tion when baseline models are already well calibrated mean of 16 years apart), approximately half of the more
(e.g., when the Hosmer-Lemeshow statistic already has than 2000 individuals studied remained within the
a P-value ⬎0.05). Reclassification has clinical relevance same quintile of CRP values (32 ).
because practice guidelines typically refer to predeter- In the past 15 years, more than 20 epidemiological
mined risk categories. The dependence of reclassifica- studies have demonstrated a significant association be-
tion metrics on the number of risk categories and the tween increased CRP concentrations as measured by
choice of cutpoints has been noted earlier. Further- use of high-sensitivity assays (hs-CRP) and the risk of a
more, reclassification metrics generally do not take first cardiovascular event among asymptomatic pa-
into account whether the shift in risk classification tients (33, 34, 35 ). A metaanalysis of 22 studies by the
would lead to a change in clinical management (16 ). US Preventive Services Task Force indicated that hs-
Thus, it has been proposed that studies addressing bio- CRP concentrations greater than or equal to 3 mg/L
markers in primary prevention should describe the were associated with a 60% increased risk of cardiovas-
performance of new biomarkers according each of cular disease (36 ). The degree to which addition of hs-
these complementary criteria (26 ). CRP measurement improves the predictive accuracy of
traditional risk stratification models, however, has
Specific Biomarkers been controversial. Studies in low-risk populations
have generally shown modest increases in the c-statistic
Over the past decade, multiple new cardiac biomarkers with the addition of hs-CRP, from 0.0 to 0.02 (37 ). The
have emerged. Below, we discuss several examples same studies have shown only modest or absent in-
of established biomarkers (CRP, BNP) and newer creases in global calibration measures. On the other
biomarkers [lipoprotein-associated phospholipase hand, investigators from the Women’s Health Study
A2 V(Lp-PLA2) and high-sensitivity troponin (hs- reported that the addition of hs-CRP reclassified a
troponin)]. We then review findings from one of the significant proportion of individuals, including ap-
newer platforms for biomarker discovery (metabolo- proximately 20% of those at “intermediate risk.”
mics). These examples are meant to be illustrative, The majority of the reclassification for intermediate-
rather than providing a comprehensive description of risk individuals (roughly 75%) were down-
available cardiovascular biomarkers. classifications into the low risk group (37 ).
The JUPITER (Justification for the Use of Statins
HIGH-SENSITIVITY CRP in Primary Prevention: An Intervention Trial Evaluat-
CRP was originally discovered by Tillett and Francis in ing Rosuvastatin) study was a large, primary preven-
1930. It was found in the serum of patients with acute tion trial with a potent statin medication (rosuvastatin)
inflammation and was noted to react with the C poly- (38 ). Its use of hs-CRP concentrations as one of the
saccharide of pneumococcus, hence its name (27 ). inclusion criteria makes it one of the few major ran-
CRP is an acute-phase reactant, produced primarily domized controlled trials in cardiology to incorporate
by hepatocytes in response to stimulation from biomarker-directed therapy. JUPITER enrolled nearly
interlukin-6 and tumor necrosis factor-␣ (28 ). Typi- 18 000 individuals (men ⬎50 and women ⬎60 years
cally, CRP binds to phosphocholine expressed on the old) with LDL concentrations less that 130 mg/dL, but
surface of dead or dying cells and activates the comple- hs-CRP concentrations ⬎2 mg/L. The study was
ment system via the C1Q complex (29 ). stopped early due to an approximately 40% reduction
CRP is a 224-residue protein with an annular, pen- in the composite cardiovascular end point with rosuv-
tameric disc shape. In general, circulating CRP concen- astatin compared with placebo. Because the study did
trations can increase up to 50 000 –fold in acute inflam- not randomize individuals with hs-CRP concentration
mation (i.e., infection) within 6 h. CRP typically peaks ⬍2 mg/L, it is uncertain whether screening with hs-
at 48 h. Because the half-life of CRP is constant, the CRP is necessary to identify older adults who would
degree of increase is determined by the severity of the benefit from statin therapy. In the 2010 American Col-
precipitating cause (30 ). Intraindividual variability of lege of Cardiology Foundation/American Heart Asso-
CRP in healthy individuals has been studied by several ciation Guideline for Assessment of Cardiovascular

Clinical Chemistry 58:1 (2012) 75

Reviews

Risk in Asymptomatic Adults, the measurement of hs- and Cancer cohort, NT-proBNP was evaluated along
CRP in older men and women with LDL ⬍130 mg/dL with CRP, lipoprotein-associated phospholipase 2,
to aid in the decision to use statins was given a class IIa midregional proadrenomedullin, and cystatin C
recommendation (39 ). Overall, measuring hs-CRP for (41 ). During more than 12 years of follow up, NT-
cardiovascular risk assessment in men ⬎50 years proBNP predicted cardiovascular and coronary
and women ⬎60 years was given a class IIb events, even after adjusting for CRP and other bio-
recommendation. markers. In the whole sample, there was only a mod-
est increase in c-statistic, and a nonsignificant NRI.
B-TYPE NATRIURETIC PEPTIDE Results were stronger when analyses were restricted
BNP is a member of a family of hormones known as the to intermediate-risk individuals, in whom signifi-
natriuretic peptides. The natriuretic peptides are syn- cant NRIs were noted (ranging from 5% to 15%).
thesized primarily in the heart, and upregulated by Most of the correct reclassification consisted of
myocardial wall stress (40 ). The gene that encodes BNP “down-classification” to lower risk categories.
[natriuretic peptide B (NPPB)3] initially produces
preproBNP. PreproBNP is cleaved to form proBNP, a LIPOPROTEIN-ASSOCIATED PHOSPHOLIPASE A2
108 –amino acid peptide precursor. Finally, proBNP is Lp-PLA2 is a 441–amino acid protein, encoded by the
cleaved by the enzyme corin, forming the biologically phospholipase A2, group VII (platelet-activating factor
active C-terminal fragment (mature BNP) as well as acetylhydrolase, plasma) (PLA2G7) gene, which is pro-
the inactive N-terminal fragment (NT-proBNP). Al- duced by inflammatory cells. It circulates mainly with
though both BNP and NT-proBNP can be measured in LDL (⬍20% is associated with HDL or remnant lipo-
plasma, the latter has a longer half-life (1–2 h, com- proteins), and it is responsible for hydrolyzing oxidized
pared with 20 min). In ambulatory populations, more phospholipids in LDL. Specifically, it catalyzes the deg-
than 90% of individuals have detectable NT-proBNP radation (hydrolysis) of platelet activating factor to in-
concentrations, compared with roughly 70% for BNP active products (47 ). It is highly upregulated in athero-
(41 ). sclerotic plaques, and may be directly involved in
Both BNP and NT-proBNP are markedly increased development of atherosclerosis (48 ) and plaque rup-
in individuals with acute heart failure (42 ), which makes ture (49 ). Because it is produced by macrophages and
them valuable biomarkers in the setting of suspected heart foam cells in the vascular intima, Lp-PLA2 is thought
failure. Interestingly, within the “normal” reference inter- to be potentially more specific for vascular inflamma-
val of values, there is significant variation in BNP among tion than other inflammatory markers, such as CRP
apparently healthy individuals, and this variation is prog- (50 ).
nostically significant (43 ). For example, in the Framing- Early in vivo studies showed that both rabbits and
ham Offspring Study, BNP concentrations ⬎20 pg/mL humans with increased concentrations of circulating
were associated with a 60%–200% increased risk of car- Lp-PLA2 have a higher plaque burden in the coronary
diovascular events, stroke, heart failure, and all-cause arteries than those with normal concentrations (51 ).
mortality during a mean follow-up of 5.2 years (45 ). A Furthermore, it has been shown experimentally that
recent metaanalysis of 40 prospective studies, approxi- inhibiting Lp-PLA2 leads to a reduction in atheroscle-
mately half of which were performed in primary preven- rotic lesions in hyperlipidemic rabbits (51 ). Given its
tion populations, found consistently strong associations low biological variation, specificity for the vasculature,
between BNP concentrations and cardiovascular risk and association with atherosclerotic plaque burden,
(Fig. 1) (46 ). Subtle increases in BNP could reflect in- Lp-PLA2 has been proposed as a potential biomarker
creased ventricular wall stress resulting from subclinical for identifying individuals at increased risk of cardio-
ischemia, high afterload, or increased neurohormonal ac- vascular disease.
tivation. Interestingly, the association between BNP and There have been more than 10 studies assessing the
future events persists even after adjustment for standard role of Lp-PLA2 in primary prevention. For instance,
echocardiographic measures (45 ). in the West of Scotland Coronary Prevention Study
There are fewer data on whether BNP measure- study, both hs-CRP and Lp-PLA2 were significantly
ments improve discrimination, calibration, or re- associated with increased cardiovascular risk (52 ). Fur-
classification for predicting cardiovascular events, thermore, after adjustment for traditional risk factors,
on top of traditional risk factors. In the Malmo Diet Lp-PLA2 maintained its association whereas hs-CRP
did not (52 ). In the Atherosclerosis Risk in Communi-
ties study, the weighted mean of both Lp-PLA2 and
hs-CRP were higher in those who went on to develop
3
Human genes: NPPB, natriuretic peptide B; PLA2G7, phospholipase A2, group coronary events than in those who did not (53 ). As with
VII (platelet-activating factor acetylhydrolase, plasma). BNP, there are fewer data regarding the impact of Lp-

76 Clinical Chemistry 58:1 (2012)

Biomarkers in Primary Prevention
Reviews

Fig. 1. Relative risks for cardiovascular disease (CVD) in individuals in the top vs bottom third of baseline BNP or
NT-proBNP concentration, according to different study characteristics.
Reproduced with permission from Di Angelantonio et al. (46 ).

PLA2 measurements on discrimination, calibration, or raised the c-statistic from 0.72 to 0.733, and improved net
reclassification. A recent nested case-control study from reclassification. In contrast, a report from the Malmo Diet
the Nurses Health Study examined discrimination with and Cancer Study suggested nonsignificant changes in
Lp-PLA2 in women (54 ). The additional of Lp-PLA2 discrimination and reclassification (41 ).

Clinical Chemistry 58:1 (2012) 77

Reviews

Fig. 2. Association of cardiac troponin T (cTnT) detected with a highly sensitive assay and all-cause and cardio-
vascular mortality.
The blue segment of the vertical axis indicates the range from 0% to 20%. Reproduced with permission from de Lemos et al.
(59 ).

HIGH-SENSITIVITY TROPONIN of the normal reference population (57 ). The hs-

In 1963, Ebashi and colleagues discovered that a new troponin assay, which can detect picomolar concentra-
myofibrillar protein, troponin, was integral to both tions of troponin, has increased the sensitivity of acute
skeletal and cardiac muscle contraction (55 ). Troponin myocardial infarction diagnosis and introduced other
has 3 subunits: troponin C, I, and T. Troponin C is potential clinical applications (58 ).
responsible for sensing and responding to the presence There have been 3 recent cohort studies examining
of calcium. Troponin T binds tropomyosin, forming the clinical utility of hs-troponins. A recent study by de
the troponin–tropomyosin complex. Troponin I binds Lemos et al. measured troponin T concentrations using
to actin in thin myofilaments, and holds the troponin– both standard and highly sensitive assays in over 3500
tropomyosin complex in place (56 ). adults in the Dallas Heart Study (59 ). The prevalence of
Cardiac troponins have become the standard bio- detectable troponin T was 25% with the high-sensitivity
marker in the diagnosis of acute myocardial infarction. assay, compared with only 0.7% for the conventional as-
On the basis of consensus guidelines, a troponin con- say. The investigators found that hs-troponin T concen-
centration above the 99th percentile (of a healthy pop- trations were associated with baseline structural heart dis-
ulation) is considered diagnostic of an acute myocar- ease (as determined based on MRI) and all-cause and
dial infarction. With conventional troponin assays, cardiovascular mortality over 6 years of follow up (Fig. 2).
troponin concentrations are undetectable in the vast This association was substantially attenuated with the use
majority of healthy individuals. However, newer gen- of conventional troponin T. The addition of hs-troponin
erations of cardiac troponin I and T assays, the so- to a multivariable-adjusted model for mortality improved
called high-sensitivity assays, permit detection of con- the c-statistic (⫹0.02, P ⫽ 0.001) and integrated discrim-
centrations significantly lower than the 99th percentile ination index (0.04, P⬍0.001).

78 Clinical Chemistry 58:1 (2012)

Biomarkers in Primary Prevention
Reviews

Fig. 3. Biomarker identification and the genome, transcriptome, proteome, and metabolome.
The numbers in the parentheses denote the estimated number of entities of each type of molecule. Reproduced with permission
from Gerszten and Wang (1 ).

deFilippi et al. studied the prognostic value of hs- which is not surprising in view of their association with
troponin T concentrations in more than 4200 older cardiac wall stress. hs-CRP and Lp-PLA2 are very good
adults (age ⬎65 years) (60 ). Increased hs-troponin T markers of vascular risk, probably reflecting the impor-
concentrations were associated with the development tance of vascular inflammation in atherogenesis (66 ).
of heart failure and cardiovascular death. The addition On the other hand, “myocardial” biomarkers such as
of hs-troponin T to traditional risk factors raised the NT-proBNP and hs-troponin are also surprisingly
c-statistics for heart failure and cardiovascular death by good predictors of future vascular events, for reasons
approximately 0.01. Similarly, investigators from the that are not entirely understood (44, 67 ).
Atherosclerosis Risk in Communities study found that
hs-troponin T concentrations were associated with in- New Directions
cident CHD, mortality, and heart failure in nearly
10000 individuals (44 ). Furthermore, the addition of Newer technologies permit the systematic study of pro-
hs-troponin T to traditional risk factors modestly im- teins (proteomics) and small molecules (metabolo-
proved discrimination and reclassification. mics) in biospecimens, including plasma or serum.
Proteomics and metabolomics have similarities to
Comparisons of Biomarkers genomics, in that they provide an unbiased approach
to biomarker identification (1 ). In contrast to genetic
A growing number of studies have provided data on the profiles, however, proteomic or metabolomic profiles
comparative performance of cardiovascular biomark- can respond to environmental influences and provide a
ers. For instance, several cohort studies suggest that “snapshot” of current physiologic states (Fig. 3).
NT-proBNP concentrations are more strongly related Metabolomics focuses on small molecules such as
to future vascular risk than hs-CRP concentrations lipids, sugars, nucleotides, and amino acids. Current
(61– 63 ). Similarly, recent studies indicate hs-troponin metabolomic platforms rely on 2 principal technolo-
measurements have predictive value comparable to gies: nuclear magnetic resonance and mass spectrome-
NT-proBNP and greater than hs-CRP (44, 59 ). Not try (MS) (68 ). The latter includes GC-MS and LC-MS.
surprisingly, panels of multiple biomarkers appear su- There are several important distinctions between
perior to individual biomarkers (61, 63, 64 ), although metabolomics and proteomics that are relevant from
the optimal composition of multimarker panels re- an analytic standpoint. The number of circulating me-
mains to be established. tabolites is unknown, but is estimated to be in the thou-
Furthermore, biomarkers may differ in their abil- sands, several orders of magnitude lower than the
ity to predict specific types of cardiovascular events. number of proteins. In addition, circulating metabo-
This variation often reflects the physiology underlying lites span a much smaller range of concentrations than
the biomarkers. For instance, BNP and NT-proBNP proteins (69 ). Metabolites typically reflect activity that
strongly predict incident heart failure risk (45, 65 ), is further downstream of gene expression and poten-

Clinical Chemistry 58:1 (2012) 79

Reviews

tially more closely related to cellular function than do diabetes (75 ). At a score threshold corresponding to a
proteins. Thus, circulating metabolites have the poten- 20% likelihood of coronary artery disease, the sensitiv-
tial advantage of providing immediate information ity and specificity were 85% and 43%. More studies are
about an organism’s physiologic condition, but the po- warranted to validate this approach, and to investigate
tential disadvantage of being affected by multiple whether profiles from specific cell populations might
sources of variability including diet, activity, and med- enable more accurate detection of occult cardiovascu-
ication use. lar disease.
An early study by Brindle and colleagues demon-
strated the promise and potential pitfalls of metabolo- Conclusion
mic approaches to cardiovascular screening (70 ).
They analyzed (1 ) H-NMR sera in individuals with Biomarkers hold the promise of earlier and more accu-
angiography-proven coronary artery disease, and com- rate cardiovascular risk stratification. Their role in
pared the profile with those in healthy controls. They acute cardiovascular diseases, such as myocardial in-
developed a profile that identified the cases with ⬎90% farction and heart failure, has been well studied. An
accuracy. However, a subsequent report showed that increasing number of studies have investigated the role
common variables, including sex and statin use, were of biomarkers in primary prevention. To date, no bio-
important confounders of the nuclear magnetic reso- marker has emerged as the best screening marker for
nance spectroscopy profile (71 ). cardiovascular disease, and it is likely that no single
Several groups have used targeted LC-MS ap- biomarker will be sufficiently sensitive or specific to be
proaches to reduce potential susceptibility to con- used on its own.
founding inherent in more global screens. For instance, Future strategies will likely involve larger bio-
Shah and colleagues used targeted LC-MS to profile 69 marker panels and more specific target populations.
metabolites in individuals with and without coronary Larger panels require new biomarkers that provide
artery disease referred to an academic cardiac catheter- nonoverlapping information to already -available bio-
ization laboratory (72 ). Two principal components markers or risk factors (16 ). Newer, “unbiased” ap-
that were analysis-derived factors, one comprising proaches, such as proteomics or metabolomics, show
branched-chain amino acids and another comprising some promise in this regard. Further studies are also
urea cycle metabolites, were associated with the pres- needed to better define the target populations for bio-
ence of coronary artery disease. Similarly, we recently marker screening, because a very broad approach may
employed targeted LC-MS to investigate predictors of be unnecessary as many individuals are adequately risk
future diabetes in the Framingham Offspring Study stratified by traditional risk factors.
(73 ). Five branched-chain and aromatic amino acids
were strongly associated with incident diabetes over 12
years of follow-up (P ⫽ 0.001 to ⬍0.0001). The find-
ings were replicated in an independent Swedish cohort. Author Contributions: All authors confirmed they have contributed to
Another recent study demonstrated the poten- the intellectual content of this paper and have met the following 3 re-
tial for metabolomic approaches to elucidate etio- quirements: (a) significant contributions to the conception and design,
logic factors in cardiovascular disease. Wang and acquisition of data, or analysis and interpretation of data; (b) drafting
colleagues used nontargeted LC-MS to profile spec- or revising the article for intellectual content; and (c) final approval of
the published article.
imens from individuals with and without incident
myocardial infarction (74 ). They identified 3 metab- Authors’ Disclosures or Potential Conflicts of Interest: Upon man-
olites derived from phosphatidylcholine— choline, uscript submission, all authors completed the Disclosures of Potential
Conflict of Interest form. Potential conflicts of interest:
betaine, and trimethylamine-N-oxide—that were
associated with incident events. Dietary supplemen- Employment or Leadership: None declared.
tation of these metabolites promoted atherosclerosis Consultant or Advisory Role: T.J. Wang, Diasorin, Inc.
Stock Ownership: None declared.
in animal models, which was inhibited by suppres- Honoraria: T.J. Wang, Roche Diagnostics and Quest Diagnostics.
sion of gut microflora involved in the metabolism of Research Funding: T.J. Wang, NIH R01-HL086875, R01-HL102780,
phosphatidylcholine. R01-DK081572, R01-HL098283, and R01-HL083197; Diasorin;
Gene expression profiling represents another rel- Brahms; Critical Diagnostics; Singulex; Siemens Healthcare Diag-
atively unbiased technology for biomarker discovery. nostics; and Roche Diagnostics.
Expert Testimony: None declared.
Rosenberg and colleagues recently derived and tested a Other: T.J. Wang, coinventor on patent applications relating to the
23-gene expression profile from whole blood for de- use of metabolomic and neurohormonal biomarkers in risk
tecting coronary artery disease in individuals without prediction.

80 Clinical Chemistry 58:1 (2012)

Biomarkers in Primary Prevention
Reviews

References
1. Gerszten RE, Wang TJ. The search for new car- Mol Med 2008;14:261–7. the American Heart Association. Circulation
diovascular biomarkers. Nature 2008;451:949 – 20. Cook NR. Use and misuse of the receiver operat- 2003;107:499 –511.
52. ing characteristic curve in risk prediction. Circu- 35. Zacho J, Tybjaerg-Hansen A, Jensen JS, Grande P,
2. Khot UN, Khot MB, Bajzer CT, Sapp SK, Ohman lation 2007;115:928 –35. Sillesen H, Nordestgaard BG. Genetically elevated
EM, Brener SJ, et al. Prevalence of conventional 21. D’Agostino RB Sr, Grundy S, Sullivan LM, Wilson C-reactive protein and ischemic vascular disease.
risk factors in patients with coronary heart dis- P. Validation of the Framingham coronary heart N Engl J Med 2008;359:1897–908.
ease. JAMA 2003;290:898 –904. disease prediction scores: results of a multiple 36. Buckley DI, Fu R, Freeman M, Rogers K, Helfand
3. Biomarkers Definitions Working Group. Biomark- ethnic groups investigation. JAMA 2001;286: M. C-reactive protein as a risk factor for coronary
ers and surrogate endpoints: preferred definitions 180 –7. heart disease: a systematic review and meta-
and conceptual framework. Clin Pharmacol Ther 22. Expert Panel on Detection, Evaluation, and Treat- analyses for the U.S. Preventive Services Task
2001;69:89 –95. ment of High Blood Cholesterol in Adults. Exec- Force. Ann Intern Med 2009;151:483–95.
4. Morrow DA, de Lemos JA. Benchmarks for the utive Summary of The Third Report of The Na- 37. Cook NR, Buring JE, Ridker PM. The effect of
assessment of novel cardiovascular biomarkers. tional Cholesterol Education Program (NCEP) including C-reactive protein in cardiovascular risk
Circulation 2007;115:949 –52. Expert Panel on Detection, Evaluation, And Treat- prediction models for women. Ann Intern Med
5. Wang TJ, Larson MG, Levy D, Leip EP, Benjamin ment of High Blood Cholesterol In Adults (Adult 2006;145:21–9.
EJ, Wilson PW, et al. Impact of age and sex on Treatment Panel III). JAMA : the journal of the 38. Ridker PM, Danielson E, Fonseca FA, Genest J,
plasma natriuretic peptide levels in healthy American Medical Association 2001;285:2486 – Gotto AM Jr, Kastelein JJ, et al. Rosuvastatin to
adults. Am J Cardiol 2002;90:254 – 8. 97. prevent vascular events in men and women with
6. Redfield MM, Rodeheffer RJ, Jacobsen SJ, Ma- 23. Wilson PW, D’Agostino RB, Levy D, Belanger AM, elevated C-reactive protein. N Engl J Med 2008;
honey DW, Bailey KR, Burnett Jr. Plasma brain Silbershatz H, Kannel WB. Prediction of coronary 359:2195–207.
natriuretic peptide concentration: impact of age heart disease using risk factor categories. Circu- 39. Greenland P, Alpert JS, Beller GA, Benjamin EJ,
and gender. J Am Coll Cardiol 2002;40:976 – 82. lation 1998;97:1837– 47. Budoff MJ, Fayad ZA, et al. 2010 ACCF/AHA
7. Lam CS, Cheng S, Choong K, Larson MG, Mu- 24. Pencina MJ, D’Agostino RB Sr, D’Agostino RB Jr, guideline for assessment of cardiovascular risk in
rabito JM, Newton-Cheh C, et al. Influence of sex Vasan RS. Evaluating the added predictive ability asymptomatic adults: a report of the American
and hormone status on circulating natriuretic of a new marker: from area under the ROC curve College of Cardiology Foundation/American Heart
peptides. J Am Coll Cardiol 2011;58:618 –26. to reclassification and beyond. Stat Med 27:157– Association Task Force on Practice Guidelines.
8. Chang AY, Abdullah SM, Jain T, Stanek HG, Das 72, 2008; discussion 207–12. J Am Coll Cardiol 2010;56:e50 –103.
SR, McGuire DK, et al. Associations among an- 25. Pencina MJ, D’Agostino RB Sr, Steyerberg EW. 40. Levin ER, Gardner DG, Samson WK. Natriuretic
drogens, estrogens, and natriuretic peptides in Extensions of net reclassification improvement peptides. N Engl J Med. 1998;339:321– 8.
young women: observations from the Dallas calculations to measure usefulness of new bio- 41. Melander O, Newton-Cheh C, Almgren P, Hed-
Heart Study. J Am Coll Cardiol 2007;49:109 –16. markers. Stat Med 2011;30:11–21. blad B, Berglund G, Engstrom G, et al. Novel and
9. Wang TJ, Larson MG, Levy D, Benjamin EJ, Leip 26. Hlatky MA, Greenland P, Arnett DK, Ballantyne conventional biomarkers for prediction of inci-
EP, Wilson PW, et al. Impact of obesity on plasma CM, Criqui MH, Elkind MS, et al. Criteria for dent cardiovascular events in the community.
natriuretic peptide levels. Circulation 2004;109: evaluation of novel markers of cardiovascular JAMA 2009;302:49 –57.
594 – 600. risk: a scientific statement from the American 42. Burnett JC Jr, Kao PC, Hu DC, Heser DW, Heublein
10. Mehra MR, Uber PA, Park MH, Scott RL, Ventura Heart Association. Circulation 2009;119:2408 – D, Granger JP, et al. Atrial natriuretic peptide
HO, Harris BC, et al. Obesity and suppressed 16. elevation in congestive heart failure in the hu-
B-type natriuretic peptide levels in heart failure. 27. Tillett WS, Francis T. Serological reactions in man. Science 1986;231:1145–7.
J Am Coll Cardiol 2004;43:1590 –5. pneumonia with a non-protein somatic fraction 43. McKie PM, Cataliotti A, Lahr BD, Martin FL, Red-
11. Das SR, Drazner MH, Dries DL, Vega GL, Stanek of pneumococcus. J Exp Med 1930;52:561–71. field MM, Bailey KR, et al. The prognostic value
HG, Abdullah SM, et al. Impact of body mass and 28. Kushner I. The phenomenon of the acute phase of N-terminal pro-B-type natriuretic peptide for
body composition on circulating levels of natri- response. Ann NY Acad Sci 1982;389:39 – 48. death and cardiovascular events in healthy nor-
uretic peptides: results from the Dallas Heart 29. Thompson D, Pepys MB, Wood SP. The physio- mal and stage A/B heart failure subjects. J Am
Study. Circulation 2005;112:2163– 8. logical structure of human C-reactive protein and Coll Cardiol 2010;55:2140 –7.
12. McCord J, Mundy BJ, Hudson MP, Maisel AS, its complex with phosphocholine. Structure 1999; 44. Saunders JT, Nambi V, de Lemos JA, Chambless
Hollander JE, Abraham WT, et al. Relationship 7:169 –77. LE, Virani SS, Boerwinkle E, et al. Cardiac tro-
between obesity and B-type natriuretic peptide 30. Pepys MB, Hirschfield GM. C-reactive protein: a ponin T measured by a highly sensitive assay
levels. Arch Intern Med 2004;164:2247–52. critical update. J Clin Invest 2003;111:1805–12. predicts coronary heart disease, heart failure, and
13. Horwich TB, Hamilton MA, Fonarow GC. B-type 31. Macy EM, Hayes TE, Tracy RP. Variability in the mortality in the Atherosclerosis Risk in Commu-
natriuretic peptide levels in obese patients with measurement of C-reactive protein in healthy nities Study. Circulation 2011;123:1367–76.
advanced heart failure. J Am Coll Cardiol 2006; subjects: implications for reference intervals and 45. Wang TJ, Larson MG, Levy D, Benjamin EJ, Leip
47:85–90. epidemiological applications. Clinical Chemistry EP, Omland T, et al. Plasma natriuretic peptide
14. Koenig W. Cardiovascular biomarkers: added 1997;43:52– 8. levels and the risk of cardiovascular events and
value with an integrated approach? Circulation 32. Chen TH, Gona P, Sutherland PA, Benjamin EJ, death. N Engl J Med 2004;350:655– 63.
2007;116:3–5. Wilson PW, Larson MG, et al. Long-term 46. Di Angelantonio E, Chowdhury R, Sarwar N, Ray
15. Ware JH. The limitations of risk factors as prog- C-reactive protein variability and prediction of KK, Gobin R, Saleheen D, et al. B-type natriuretic
nostic tools. N Engl J Med 2006;355:2615–7. metabolic risk. Am J Med 2009;122:53– 61. peptides and cardiovascular risk: systematic re-
16. Wang TJ. Assessing the role of circulating, ge- 33. Ridker PM, Cushman M, Stampfer MJ, Tracy RP, view and meta-analysis of 40 prospective studies.
netic, and imaging biomarkers in cardiovascular Hennekens CH. Inflammation, aspirin, and the Circulation 2009;120:2177– 87.
risk prediction. Circulation 2011;123:551– 65. risk of cardiovascular disease in apparently 47. O’Donoghue M, Morrow DA, Sabatine MS, Mur-
17. Hanley JA, McNeil BJ. The meaning and use of healthy men. N Engl J Med 1997;336:973–9. phy SA, McCabe CH, Cannon CP, et al.
the area under a receiver operating characteristic 34. Pearson TA, Mensah GA, Alexander RW, Ander- Lipoprotein-associated phospholipase A2 and its
(ROC) curve. Radiology 1982;143:29 –36. son JL, Cannon RO 3rd, Criqui M, et al. Markers association with cardiovascular outcomes in pa-
18. Wilson PW. Prediction of cardiovascular disease of inflammation and cardiovascular disease: ap- tients with acute coronary syndromes in the
events. Cardiol Clin 2011;29:1–13. plication to clinical and public health practice: A PROVE IT-TIMI 22 (PRavastatin Or atorVastatin
19. May A, Wang TJ. Biomarkers for cardiovascular statement for healthcare professionals from the Evaluation and Infection Therapy-Thrombolysis In
disease: challenges and future directions. Trends Centers for Disease Control and Prevention and Myocardial Infarction) trial. Circulation 2006;113:

Clinical Chemistry 58:1 (2012) 81

Reviews

1745–52. 57. Christenson RH, Phillips D. Sensitive and high 66. Ridker PM, Hennekens CH, Buring JE, Rifai N.
48. Zalewski A, Macphee C. Role of lipoprotein- sensitivity next generation cardiac troponin C-reactive protein and other markers of inflam-
associated phospholipase A2 in atherosclerosis: assays: more than just a name. Pathology 2011; mation in the prediction of cardiovascular disease
biology, epidemiology, and possible therapeutic 43:213–9. in women. N Engl J Med 2000;342:836 – 43.
target. Arterioscler Thromb Vasc Biol 2005;25: 58. Scirica BM. Acute coronary syndrome: emerging 67. Emberson JR, Ng LL, Armitage J, Bowman L,
923–31. tools for diagnosis and risk assessment. J Am Coll Parish S, Collins R. N-terminal Pro-B-type natri-
49. Anderson JL. Lipoprotein-associated phospho- Cardiol 2010;55:1403–15. uretic peptide, vascular disease risk, and choles-
lipase A2: an independent predictor of coronary 59. de Lemos JA, Drazner MH, Omland T, Ayers CR, terol reduction among 20,536 patients in the
artery disease events in primary and secondary Khera A, Rohatgi A, et al. Association of troponin MRC/BHF heart protection study. J Am Coll Car-
prevention. Am J Cardiol 2008;101:23F–33F. T detected with a highly sensitive assay and diol 2007;49:311–9.
50. McConnell JP, Hoefner DM. Lipoprotein- cardiac structure and mortality risk in the general 68. Mayr M. Metabolomics: ready for the prime time?
associated phospholipase A2. Clin Lab Med 2006; population. JAMA 2010;304:2503–12. Circ Cardiovasc Genet 2008;1:58 – 65.
26:679 –97, vii. 60. deFilippi CR, de Lemos JA, Christenson RH, Gott- 69. Wishart DS. Human Metabolome Database: com-
51. Hakkinen T, Luoma JS, Hiltunen MO, Macphee diener JS, Kop WJ, Zhan M, et al. Association of pleting the “human parts list.” Pharmacogenom-
CH, Milliner KJ, Patel L, et al. Lipoprotein- serial measures of cardiac troponin T using a ics 2007;8:683– 6.
associated phospholipase A(2), platelet- sensitive assay with incident heart failure and 70. Brindle JT, Antti H, Holmes E, Tranter G, Nichol-
activating factor acetylhydrolase, is expressed by cardiovascular mortality in older adults. JAMA son JK, Bethell HW, et al. Rapid and noninvasive
macrophages in human and rabbit atherosclerotic 2010;304:2494 –502.
diagnosis of the presence and severity of coro-
lesions. Arterioscler Thromb Vasc Biol 1999;19: 61. Wang TJ, Gona P, Larson MG, Tofler GH, Levy D,
nary heart disease using 1H-NMR-based metabo-
2909 –17. Newton-Cheh C, et al. Multiple biomarkers for
nomics. Nat Med 2002;8:1439 – 44.
52. Packard CJ, O’Reilly DS, Caslake MJ, McMahon the prediction of first major cardiovascular events
71. Kirschenlohr HL, Griffin JL, Clarke SC, Rhydwen R,
AD, Ford I, Cooney J, et al. Lipoprotein-associated and death. N Engl J Med 2006;355:2631–9.
Grace AA, Schofield PM, et al. Proton NMR anal-
phospholipase A2 as an independent predictor of 62. Wannamethee SG, Welsh P, Lowe GD, Gudnason
ysis of plasma is a weak predictor of coronary
coronary heart disease. West of Scotland Coro- V, Di Angelantonio E, Lennon L, et al. N-terminal
artery disease. Nat Med 2006;12:705–10.
nary Prevention Study Group. N Engl J Med 2000; pro-brain natriuretic Peptide is a more useful
72. Shah SH, Bain JR, Muehlbauer MJ, Stevens RD,
343:1148 –55. predictor of cardiovascular disease risk than
Crosslin DR, Haynes C, et al. Association of a
53. Ballantyne CM, Hoogeveen RC, Bang H, Coresh J, C-reactive protein in older men with and without
Folsom AR, Heiss G, et al. Lipoprotein-associated pre-existing cardiovascular disease. J Am Coll peripheral blood metabolic profile with coronary
phospholipase A2, high-sensitivity C-reactive pro- Cardiol 2011;58:56 – 64. artery disease and risk of subsequent cardiovas-
tein, and risk for incident coronary heart disease 63. Zethelius B, Berglund L, Sundstrom J, Ingelsson E, cular events. Circ Cardiovasc Genet 2010;3:207–
in middle-aged men and women in the Athero- Basu S, Larsson A, et al. Use of multiple biomark- 14.
sclerosis Risk in Communities (ARIC) study. Cir- ers to improve the prediction of death from car- 73. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee
culation 2004;109:837– 42. diovascular causes. N Engl J Med 2008;358: EP, McCabe E, et al. Metabolite profiles and the
54. Hatoum IJ, Cook NR, Nelson JJ, Rexrode KM, 2107–16. risk of developing diabetes. Nat Med 2011;17:
Rimm EB. Lipoprotein-associated phospholipase 64. Blankenberg S, Zeller T, Saarela O, Havulinna AS, 448 –53.
A2 activity improves risk discrimination of inci- Kee F, Tunstall-Pedoe H, et al. Contribution of 30 74. Wang Z, Klipfell E, Bennett BJ, Koeth R, Levison
dent coronary heart disease among women. Am biomarkers to 10-year cardiovascular risk estima- BS, Dugar B, et al. Gut flora metabolism of phos-
Heart J 2011;161:516 –22. tion in 2 population cohorts: the MONICA, risk, phatidylcholine promotes cardiovascular disease.
55. Ebashi S. Third component participating in the genetics, archiving, and monograph (MORGAM) Nature 2011;472:57– 63.
superprecipitation of ‘natural actomyosin’. Na- biomarker project. Circulation 2010;121:2388 – 75. Rosenberg S, Elashoff MR, Beineke P, Daniels SE,
ture 1963;200:1010. 97. Wingrove JA, Tingley WG, et al. Multicenter val-
56. Rybakova IN, Greaser ML, Moss RL. Myosin bind- 65. Velagaleti RS, Gona P, Larson MG, Wang TJ, Levy idation of the diagnostic accuracy of a blood-
ing protein C interaction with actin: characteriza- D, Benjamin EJ, et al. Multimarker approach for based gene expression test for assessing obstruc-
tion and mapping of the binding site. J Biol Chem the prediction of heart failure incidence in the tive coronary artery disease in nondiabetic
2011;286:2008 –16. community. Circulation 2010;122:1700 – 6. patients. Ann Intern Med 2010;153:425–34.

82 Clinical Chemistry 58:1 (2012)