Update on integrated biomarkers for assessment of long-term risk of cardiovascular complications in initially healthy subjects and patients with manifest atherosclerosis

Abstract

Risk stratification for cardiovascular diseases (CVD) remains suboptimal even after the introduction of global risk assessment by various scores. This has prompted the search for additional biomarkers which might help to improve risk stratification. Basically, there are blood biomarkers representing various pathophysiological pathways of atherosclerosis, and markers of subclinical disease. Since inflammatory processes accompany all stages of atherosclerosis, measurement of plasma/serum concentrations of circulating inflammatory biomarkers have received great attention. Such biomarkers can be measured systemically by sensitive assays, and elevated concentrations in the circulation have been shown to be associated with future CVD events. Thus, they might add to the predictive value of the atherogenic lipoprotein phenotype to further improve CVD risk assessment. In addition, several non-invasive imaging techniques are available for which also a predictive value for CVD could be established. However, for most of these biomarkers the clinical utility has not yet been firmly established. This review attempts to give an update on the potential use of biomarkers for risk stratification in initially healthy subjects and patients with manifest, chronic atherosclerosis, particularly focusing on the integrated value of the combination of these markers.

• Biomarker
• cardiovascular risk assessment
• integrated approach

Introduction

In primary prevention, traditional risk factors are a useful first step in determining who is at cardiovascular risk. In the era of ‘global risk assessment’ scores, e.g. the Framingham Score [1], the PROCAM Score [2], or the European Society of Cardiology SCORE [3], derived from multivariable statistical models should be used. However, it has been noted that a considerable number of those at risk cannot be identified on the basis of traditional risk factors alone [4–6]. This has prompted the search for novel markers of cardiovascular risk. Such markers could either represent blood biomarkers relevant to the pathophysiology of atherosclerosis, e.g. representing inflammatory pathways, coagulation, platelet aggregation, lipoproteins or lipid-related variables, genetic markers, or markers of subclinical disease which also may aid in improved risk prediction. Subjects at intermediate risk (10-year risk of 10%–20%) would represent candidates for additional testing, to increase or decrease their actual risk [7].

In patients with manifest chronic atherosclerotic disease there is also on-going research for refinement of prognosis by using some of the emerging biomarkers.

This review attempts to give an update on the potential use of biomarkers for risk stratification in initially healthy subjects and patients with manifest, chronic atherosclerosis, particularly focusing on the integrated value of the combination of these markers.

Key messages

• Risk stratification for cardiovascular diseases remains suboptimal even after the introduction of global risk assessment by various scores.

• Various biomarkers, like blood biomarkers or markers of subclinical disease, may aid in improving risk assessment, especially in those at intermediate risk for future events (10%–20% per 10 years).

• However, for most of these biomarkers the clinical utility has not yet been firmly established, and an additive value of their combination is also unclear at present.

Blood biomarkers

An important feature of atherosclerosis consists of a non-specific local inflammatory process which is accompanied by a low-grade systemic response. Thus, a number of prospective studies in initially healthy subjects have convincingly demonstrated a strong and independent association between even slightly elevated concentrations of various systemic markers of inflammation and important cardiovascular end-points (reviewed in [8]). Presently, the largest database exists for CRP, the classical acute-phase protein [9], [10]. The measurement procedure is well standardized and automated, and high-sensitivity assays with sufficient precision are available [11]. Based on substantial evidence of a contribution of inflammation to atherothrombogenesis, the recent American Heart Association/Center for Disease Control and Prevention (AHA/CDC) consensus report recommends the measurement of CRP in asymptomatic subjects at intermediate risk for future coronary events at the discretion of the physician (10-year risk of 10%–20%; class IIa, level of evidence B) [12].

Abbreviations

Table

ABI	ankle-brachial index
ACS	acute coronary syndrome
AHA	American Heart Association
ApoB	apolipoprotein B
ARIC	Atherosclerosis Risk In Communities
AUC	area under the curve
BNP	brain natriuretic peptide
CAC	coronary artery calcium
CDC	Centers for Disease Control
CHD	coronary heart disease
CHS	Cardiovascular Health Study
CI	confidence interval
CRP	C-reactive protein
CVD	cardiovascular disease
EBCT	electronic beam computed tomography
FRS	Framingham risk score
HOPE	Heart Outcomes Prevention Evaluation
HR	hazard ratio
ICAM	soluble intercellular adhesion molecules
VCAM	soluble vascular cell adhesion molecules
IL	Interleukin
IMT	intima-media thickness
KORA	Kooperative Gesundheitsforschung im Raum Augsburg
LDL	low-density lipoprotein
Lp-PLA2	lipoprotein-associated phospholipase A2
LURIC	Ludwigshafen Risk and Cardiovascular Health Study
MCP	monocyte chemoattractant protein
MDCT	multidetector computed tomography
MESA	Multi-Ethnic Study of Atherosclerosis
MI	myocardial infarction
MONICA	MONItoring trends and determinants in CArdiovascular disease
MMP	matrix metalloproteinase
MPO	myeloperoxidase
MSCT	multislice computed tomography
NT-proBNP	N-terminal pro-brain natriuretic peptide
PAI	plasminogen activator inhibitor
PROCAM	Prospective Cardiovascular Münster
PWV	pulse wave velocity
ROC	receiver operation characteristic
SAA	serum amyloid A
SAP	serum amyloid P component
SBHW	South Bay Heart Watch
SCORE	systematic coronary risk evaluation
SD	standard deviation
SFHS	St. Francis Heart Study
SNP	single nucleotide polymorphism
sPLA2	secretory phospholipase A2
t-PA	tissue plasminogen activator
ULSAM	Uppsala Longitudinal Study of Adult Men
WHS	Women's Health Study
WOSCOPS	West of Scotland Primary Prevention Study

Atherosclerosis is also associated with disorders of the coagulation system, in particular in the ACS. A database at least as large as for CRP exists for fibrinogen, which exerts a dual role, namely as a central component in the coagulation cascade and as an acute-phase protein. A consistent and strong association between elevated fibrinogen concentrations (>300 mg/dL) and various cardiovascular outcomes, but also non-cardiovascular diseases, like cancer and total mortality, has been shown [13].

However, there are other emerging biomarkers, like Lp-PLA₂, an enzyme produced by monocytes/macrophages, T-cells, and mast cells, which has been found to generate proinflammatory and proatherogenic molecules from oxidized (ox) LDL cholesterol [14]. Further markers for risk stratification in initially healthy subjects of recent interest include other acute-phase reactants like SAA, SAP, acutephase proteins with a role in coagulation like PAI-1 and D-dimer, several cytokines like IL-6 and 18, ox LDL, other phospholipases, like type II sPLA2, MPO, various MMP, MCP-1, adipocytokines, and others [8].

Combination of blood biomarkers

Several studies have assessed the additive value of various blood biomarkers that have already proven to be predictive for future cardiovascular events. Mora et al. [15] prospectively assessed the combined effect of fibrinogen and CRP in the prediction of incident cardiovascular events in the WHS showing an age-adjusted hazard ratio of 3.45 in those with fibrinogen in the top tertile and CRP > 3 mg/L compared to those in the bottom tertile of each parameter. Subjects with either fibrinogen levels in the top tertile or CRP > 3 mg/L showed an almost identical but intermediate risk. High CRP and high D-dimer levels clearly showed an additive effect on the risk of CHD in the Caerphilly and Speedwell studies [16]. An even larger panel of inflammation-sensitive plasma proteins was investigated in the Malmö study [17]. Here, fibrinogen, haptoglobin, α₁-antitrypsin, caeruloplasmin, and orosomucoid were measured in addition to total cholesterol in more than 6,000 healthy men aged 28–61 and followed for almost 19 years. Basically, each of the inflammation-sensitive proteins added information to total cholesterol, but, even more impressive, the number of positive (> than median) inflammation sensitive-proteins was clearly related to fatal and non-fatal cardiovascular events and to all-cause mortality, however less strongly. In the Edinburgh Artery Study [18] in almost 1,600 men and women aged 55–64 and followed for 17 years, 17 biomarkers of inflammation, haemostasis, and blood rheology were related to incident CHD and stroke. However, although IL-6, fibrinogen, tissue-type plasminogen activator (t-PA), and ICAM-1 were significantly related to outcome in multivariable analyses, no significant increase was seen in the AUC in ROC analyses for any of the 17 biomarkers, in addition to traditional risk factors plus ankle-brachial index (ABI) as a marker of subclinical disease. However, cardiovascular disease risk increased with the number of elevated markers. Finally, concerning inflammation, the addition of CRP to the metabolic syndrome has been shown to improve risk prediction in the WOSCOPS study [19] as well as in the WHS [20]. In subjects with metabolic syndrome and CRP ≥ 3 mg/L the risk for either CHD events or onset of new diabetes was clearly elevated compared to subjects with CRP < 3 mg/L.

Since Lp-PLA₂ does not correlate with most other risk factors, an additive effect of CRP and Lp-PLA₂ is plausible which indeed has been demonstrated in the MONICA/KORA Augsburg study for incident CHD in middle-aged men during a 14-year follow-up [21] as well as for stroke in the ARIC study [22]. In the large LURIC study [23], Lp-PLA₂ added prognostic information in patients with low and medium high-sensitivity (hs) CRP with regard to 5-year cardiac mortality independently of established risk factors. Finally, data from the Bruneck study [24] in middle-aged men and women showed an additive effect of risk prediction for the combination of lipoprotein (a) (Lp(a)) and Lp-PLA₂ as well as the ratio of oxidized phospholipids/apo B and Lp-PLA_2.

Investigating the predictive value of low adiponectin for the risk of type 2 diabetes and CHD events in apparently healthy middle-aged men from the Augsburg MONICA/KORA study [25], it could be shown that subjects with low adiponectin and low high-density lipoprotein (HDL) cholesterol were of particularly high risk for incident type 2 diabetes and CHD events. Such increased risk for cardiometabolic outcomes may also apply to other combinations of various biomarkers, although the evidence so far is limited.

Most importantly, to date, it has not been shown unequivocally that the addition of any of the emerging biomarkers to traditional risk factors provides incremental information above and beyond that contained in global scoring using classical cardiovascular risk factors. A prominent example of such an exercise comes from the Framingham Heart Study [26], in which 10 biomarkers related to cardiovascular disease were studied in more than 3,200 subjects followed for a median of 7.4 years. Biomarkers studied included CRP, natriuretic peptides such as BNP, N-terminal pro-atrial natriuretic peptide (NT-proANP), fibrinogen, D-dimer, PAI-1, homocysteine, renin, and the urinary albumin-to-creatinine ratio. The biomarkers that most strongly predicted major cardiovascular events were BNP and the urinary albumin-to-creatinine ratio, and a high value in the multimarker score was clearly predictive for death or cardiovascular events. However, the addition of the multimarker score to conventional risk factors resulted in only minimal increases in the ability to classify risk as measured by C-statistics. This study has been discussed controversially for the modest number of end-points, which in addition contained a combination of hard outcomes, like fatal and non-fatal MI, as well as weaker end-points, i.e. angina pectoris and congestive heart failure.

More recently Zethelius et al. [27] have evaluated the contribution of multiple biomarkers for the prediction of hard outcomes such as death from all causes and from CVD over a 10-year follow-up period. Four biomarkers, related to cardiac and renal disease as well as inflammation, such as troponin I, NT-proBNP, cystatin C, and CRP, were measured within the ULSAM, a community-based cohort of 1,135 elderly men, aged 71 years at base-line, of whom 661 subjects were without prevalent CVD at base-line. During a follow-up of 10 years, 315 participants died; 136 deaths were from CVD. In this analysis, an increase of 1 SD in the concentration of each biomarker, if taken separately, was significantly associated with death from CVD and all causes after multivariable adjustment, and the magnitude of the associations was almost the same within the whole cohort and after exclusion of subjects with prevalent CVD at base-line. Using state-of-the-art statistical tools, the authors could further show an improvement in risk stratification for future cardiac and total mortality by adding biomarkers to a model that included established risk factors, both in the whole sample and in participants without CVD at base-line. Furthermore, in the subgroup of subjects without prevalent disease at base-line, the addition of all four biomarkers to the established risk factors resulted in 133 appropriate and in 69 inappropriate risk reclassifications. The clinical implications of such reclassifications, however, remain unclear at present [28]. Finally, model calibration and global fit also confirmed the incremental value of the combination of biomarkers beyond conventionally available cardiovascular risk factors. Thus, the strong independent associations seen in this study and the clear incremental value of several blood biomarkers in addition to the Framingham risk score suggest that these markers may be more relevant for fatal outcomes than for non-fatal end-points.

Blankenberg [29] et al. evaluated a multimarker approach for CVD risk in patients with manifest atherosclerosis in the OPE study. The authors measured 10 biomarkers related to various pathophysiologic pathways in 3,199 patients with stable CHD, followed for 4.5 years for recurrent cardiovascular events. In Cox regression analysis, including traditional risk factors and several biomarkers, only four of them (NT-proBNP, sICAM-1 , soluble IL-1 receptor antagonist, and fibrinogen) remained significantly associated with the primary outcome. However, inclusion of these four biomarkers to a basic model consisting of simple traditional risk factors did not further improve model accuracy, which was already achieved by adding NT-proBNP (increase of AUC from 0.65 to 0.71; P <0.001).

Thus, taking into account these controversial results, further studies are needed to prove or disprove the incremental value of emerging biomarkers in addition to traditional risk profiles and the potential value of combining various blood biomarkers for risk prediction in initially healthy subjects and those with manifest atherosclerosis. An overview of studies discussed here can be found in Table I.

Table I.  Integration of multiple blood biomarker for risk assessment of cardiovascular disease (CVD).

Authors, year (reference)	Design	Cohort	n	FU	End-point	Biomarkers	Risk estimates (95% confidence interval)
Mora et al., 2006 [15]	Full cohort	Women's Health Study: healthy F	27,742	10 yrs	Incident CVD: non-fatal MI, non-fatal ischaemic stroke, coronary revascularization, CV death	CRP, Fib	HR 3.45 (2.60–4.57) for Fib >393 mg/dL and CRP >3 mg/L age-adjusted
Lowe et al., 2004 [16]	Full cohort	Caerphilly and Speedwell studies: healthy M	2,398/2,055	105 Mo/75 Mo	Incident IHD	CRP, D-dimer	IHD Incidence for CRP and D-dimer in T3 19.5% versus 7% in T1 for both BM; CRP T3: 8.9%; D-dimer T3: 8%
Engström et al., 2002 [17]	Full cohort	Malmö: healthy M/F	6,063	18.7 yrs	Incident CVD: stroke, cardiac events (fatal/non-fatal), CV deaths	Fib, haptoglobin, α1-antitrypsin, caeruloplasmin, orosomucoid	Hypercholesterolaemia and high ISP^a: Cardiac events: RR 2.3 (1.8–3.0) CV death: RR 2.4 (1.8–3.3) Ischaemic stroke: RR 2.1 (1.4–3.3) multivariate adjusted^b
Tzoulaki et al., 2007 [18]	Full cohort	Edinburgh Artery Study: healthy M	1,592	17 yrs	Incident CVD: fatal/non-fatal MI, stroke	CRP, IL-6, leukocyte elastase, Fib, D-dimer, t-PA, vWF, PV, FVII, E-Sel, VCAM-1, ICAM-1, Lp(a), urinary FpA, F1 + 2	HR 2.50 (1.58–3.97) ≥ 3 elevated BM in T3 (IL-6, Fib, t-PA, ICAM-1) multivariate adjusted^b
Sattar et al., 2003 [19]	Full cohort	WOSCOPS: hyper-cholesterolaemic M	6,3447	4.9 yrs	Incident CHD	TC, TG, LDL-C, HDL-C, fasting glucose	↑CRP/ − MS: HR 1.6 (1.3–2.1) ↓CRP/ + MS: HR 1.6 (1.2–2.1) ↑CRP/ + MS: HR 2.75 (2.1–3.6)^c
Ridker et al., 2003 [20]	Full cohort	Women's Health Study: healthy F	14,719	8 yrs	Incident CVD: MI, coronary revascularization, stroke, CV death	TC, TG, LDL-C, HDL-C, fasting glucose	↑CRP/ − MS: RR 1.5 (1.0–2.2) ↓CRP/ + MS: RR 2.3 (1.6–3.3) ↑CRP/ + MS: RR 4.0 (3.0–5.4)^c age-adjusted
Koenig et al., 2004 [21]	Full cohort	MONICA/KORA Augsburg: healthy M	934	14 yrs	Coronary events: fatal/non-fatal coronary events, incl. SCD	CRP, Lp-PLA2	HR 1.93 (1.09–3.40) for ↑CRP^c /Lp-PLA2 >290.8 ng/mL multivariate adjusted^b
Ballantyne et al., 2005 [22]	Nested case-cohort	ARIC: healthy M/F	194/766	4.4 yrs	Ischaemic stroke	CRP, Lp-PLA2	HR 11.38 (3.13–41.41) for ↑CRP^c/Lp-PLA2 in T3 multivariate adjusted^b
Winkler et al., 2007 [23]	Nested case-control	LURIC: CHD patients/controls	2513/719	5 yrs	Cardiac mortality	CRP, Lp-PLA2	HR 1.93 (1.04–3.59) for ↑CRP^c+Lp-PLA2 in T3 multivariate adjusted
Kiechl et al., 2007 [24]	Full cohort	Bruneck study: healthy M/F	765	10 yrs	Incident CVD: CV death, MI, stroke, TIA	Lp(a), Lp-PLA2, oxPL/ApoB	HR 3.9 for T1 oxPL/ApoB + T3 Lp-PLA2 HR 3.5 for T3 Lp(a)/T3 Lp-PLA2
Koenig et al., 2006 [25]	Full cohort	MONICA/KORA Augsburg: healthy M	937	18 yrs	Coronary events: fatal/non-fatal coronary events, incl. SCD	Adiponectin, HDL	HR 1.91(1.20—3.04) for low HDL/low adiponectin^d multivariate adjusted^b
Wang et al., 2006 [26]	Full cohort	Framingham Heart Study: healthy M/F	3,209	7.4 yrs	All-cause mortality, MACE	CRP, BNP, NT-proANP, aldosterone, renin, Fib, D-dimer, PAI-1, homocysteine, albumin/creatinine	High ‘multimarker’ score^e: for death: HR 4.08 (2.51–6.62), for MACE: HR 1.84 (1.11–3.05) multivariate adjusted^b C-statistics for death: from 0.80 (CV RF) to 0.88 (+BM) C statistics for MACE: from 0.76 (CV RF) to 0.77 (+BM)
Zethelius et al., 2008 [27]	Full cohort	ULSAM: elderly M	1,135	10 yrs	All-cause mortality, CVD mortality	Troponin I, NT-proBNP, cystatin C, CRP	C-statistics for CVD mortality: from 0.664 (CV RF) to 0.766 (+BM)
Blankenberg et al., 2006 [29]	Full cohort	HOPE: CHD patients	3,199	4.5 yrs	MI, stroke, CV death	CRP, Fib, IL-6, IL-18, sTNFR-1/2, sIL-1RA, sVCAM-1, sICAM-1, NT-proBNP, microalbuminuria	HR 2.50 (1.52–4.12) for NT-proBNP, Fib, sICAM-1, sIL-1RA in their T3 C-statistics for CVD: from 0.66 (CV RF) to 0.71 (+BM)

^aHigh ISP level was defined as 2–5 ISP in the top quartile.

^bMultivariate adjusted for various traditional cardiovascular risk factors.

^c↑CRP = ≥ 3 mg/L; ↓CRP = < 3 mg/L.

^dT1 adiponectin/T1 HDL versus T2 + T3 adiponectin/T2 + T3 HDL.

^eThe lowest two quintiles are labelled low risk, the third and fourth quintiles are labelled intermediate risk, and the top quintile is labelled high risk.

ApoB = apolipoprotein B; ARIC = Atherosclerosis Risk in Communities; BM = biomarker; BNP = brain natriuretic peptide; CHD = coronary heart disease; CRP = C-reactive protein; CV = cardiovascular; CVD = cardiovascular disease; CV RF = cardiovascular risk factors; E-Sel = E-selectin; F = female; F1 + 2 = prothrombin fragment 1 + 2; Fib = fibrinogen; FU = follow-up; FVII = factor VII; HDL-C = high-density lipoprotein-cholesterol; HOPE = Heart Outcomes Prevention Evaluation; HR = hazard ratio; ICAM = intercellular adhesion molecule-1; IHD = ischaemic heart disease; IL = interleukin; ISP = inflammation-sensitive plasma proteins; LDL-C = low-density lipoprotein cholesterol; Lp(a) = lipoprotein (a); Lp-PLA₂=lipoprotein-associated phospholipase A₂; LURIC = LUdwigshafen RIsk and Cardiovascular Health; M = male; MACE = major adverse coronary events; MI = myocardial infarction; Mo = months; MONICA/KORA = Monitoring of Trends and Determinants in Cardiovascular Disease/Kooperative Gesundheitsforschung in der Region Augsburg/Cooperative Health Research in the Region of Augsburg; MS = metabolic syndrome; NT-proANP = N-terminal pro-atrial natriuretic peptide; NT-proBNP = N-terminal pro-brain natriuretic peptide; oxPL = oxidized phospholipids; PAI-1 = plasminogen activator inhibitor-1; PV = plasma viscosity; RR = relative risk; SCD = sudden cardiac death; sICAM-1 = soluble intercellular adhesion molecule-1; sIL-1RA = soluble IL-1 receptor antagonist; sTNFR = soluble tumour necrosis factor receptor; T = tertile; TC = total cholesterol; TG = triglycerides; TIA = transient ischaemic attack; t-PA = tissue plasminogen activator; ULSAM = Uppsala Longitudinal Study of Adult Men; Urinary FpA = urinary fibrinopeptide A; VCAM-1 = vascular cell adhesion molecule-1; vWF = von Willebrand factor; WOSCOPS = West of Scotland Coronary Prevention Study.

Combination of genetic biomarkers and blood biomarkers

Another very promising and attractive approach, which has gained a great deal of attention lately, represents a combination of genetic and blood biomarkers. To date, several lines of evidence have suggested that circulating levels of biomarkers might be modulated by genetic variations (so-called SNP) within the genes coding for these proteins or lying within the common pathophysiological pathways. However, these SNPs taken separately might exert only a weak influence on the disease, whereas a combination of several alleles in different loci, acting simultaneously, might provide important additional information for better CVD risk stratification. Indeed, Kathiresan et al. [30] sought to investigate whether a combination of SNPs that are associated with LDL or HDL cholesterol might contribute to the risk of CVD. In the cardiovascular cohort of the Malmö Diet and Cancer Study, (n =5,414; 238 incident CVD events defined as MI, stroke, or death from CHD during 10.6-year follow-up) the authors generated a genotype score, consisting of 11 validated SNPs in 9 lipid-related genes (APOB, PCSK, LDLR, CETP, LIPC, LPL, APOE, HMGCR, ABCA1). A statistically significant increase in LDL and a decrease in HDL cholesterol were observed with increasing genotype scores. Furthermore, each copy of an unfavourable (i.e. associated either with increased LDL or with decreased HDL concentrations) allele was associated with the first cardiovascular event ratio (HR 1.15; 95% confidence interval (CI) 1.07–1.24 after multivariable analysis). However, the genotype score failed to improve CVD risk prediction over and beyond established risk factors, with an identical AUC in the risk factors model as well as in the model additionally containing the genotype score. However, when a risk reclassification procedure was applied, 26% of study participants in the intermediate risk category were reclassified into a higher or lower category when the genotype score was added to a model without. Thus, the SNP panel investigated in this study was independently predictive for future CVD outcome as well as being able to improve, although modestly, clinical risk reclassification for individual subjects beyond established risk factors [30].

Markers of subclinical disease

Various imaging methods have been applied to improve cardiovascular risk prediction (Table II). Among them, the most important are coronary calcium scoring by EBCT, which more recently has been substituted by MDCT, and the measurement of IMT by high-resolution carotid ultrasonography. In addition, simple measurement of ABI has repeatedly shown an excellent correlation with atherothrombotic burden.

Table II.  Established and emerging modalities for measuring subclinical cardiovascular disease.

Technique	Measure	Availability	Incremental predictive value over office-based risk assessment	Evidence showing relation to clinical events	Reproducibility
Carotid ultrasound	Intima-media thickness (IMT)	Widely available	Yes	Excellent	Good
Computer tomography	Coronary calcium score, volume	Moderately electronic beam computed tomography (EBCT) to widely multisilice computed tomography (MSCT) available	Some data	Excellent	Very good
Ankle-brachial index (ABI)	ABI ratio	Widely available (office-based)	Yes	Excellent	Very good
Magnetic resonance imaging	Plaque cross-sectional area, plaque components	Limited availability of centres assessing atherosclerotic plaque	No data	No data	Very good to excellent
Brachial ultrasound	Flow-mediated dilatation	Widely available but requires special training	No data	Good	Fair
Arterial stiffness	Aortic pulse wave velocity, carotid distensibility	Widely available	Some data	Moderately strong	Good

EBCT = electronic beam computed tomography; MSCT = multislice computed tomography.

Measurement of coronary calcium

Prospective data from the SBHW [31], the SFHS [32], and others have clearly shown that the presence and extent of coronary calcium is strongly associated with non-fatal and fatal CHD independently of traditional risk factors, and more recently data from a large registry have also shown that it adds to the prediction of all-cause mortality [33]. In some of these studies, in particular the SFHS, the addition of calcium scoring to global risk assessment by the Framingham algorithm clearly increased the AUC from 0.71 to 0.81. Of note, in women considered to be at low risk based on the Framingham score, recent data from MESA [34] showed that in particular the presence of advanced calcium identified a subgroup at high risk of future CHD. Finally, another more recent publication from MESA [35], of 6,722 initially healthy subjects who were followed for incident coronary events during a median follow-up of 3.8 years, demonstrated a strong and independent association between increasing coronary calcium and any coronary event, with HRs of 3.6, 7.7, and 9.7 in those with Agatson scores 1–100, 101–300, and >300, respectively. However, despite such strong associations, the AUC in particular in Caucasians showed no incremental value over and above the FRS. Thus, the latest AHA Scientific Statement concluded that EBCT or MDCT may be reasonable in measuring atherosclerosis burden in clinically selected intermediate-risk patients for refined clinical risk prediction and to select patients for more aggressive target values for lipid-lowering therapies (class IIb, level of evidence B) [36].

Measurement of intima-media thickness of the carotid artery

Assessment of carotid IMT by means of high-resolution ultrasound has also been used in a number of prospective studies during recent years, and the vast majority concludes that increased IMT is positively related to cardiovascular disease outcomes [37]. However, in several of such prospective studies the relation was only modest in nature, and still to date only one study in dyslipidaemic patients [38] suggests an incremental value of IMT measurement over and above traditional risk factors; and also only one study has provided evidence on the relation of change of IMT measurements and future cardiovascular events [39], [40].

Measurement of arterial stiffness

Another tool to assess the arterial vasculature non-invasively is arterial stiffness. This can be measured as PWV or carotid distensibility. PWV depends on structural changes that alter stiffness or elastic modules of the vascular wall which may increase with age and is also depending on calibre increases. A few studies have prospectively assessed the predictive value of arterial stiffness for CVD outcomes and total mortality, showing fairly strong associations with some end-points, at least for PWV. However, the incremental value over and above traditional risk factors was modest at best, thus questioning the clinical usefulness of this measurement at present [41–44].

Measurement of ankle-brachial index

Finally, the simplest non-invasive test for the presence of atherosclerotic burden, measurement of the ABI, has also been shown to predict cardiovascular events in a number of studies [45–47]. Because of the low costs involved and the simplicity of the test that can be done by a technician in general practice it has been recommended for wide-spread screening. However, data from the Edinburgh Artery Study [48], although showing an independent association between low ABI and fatal MI, only demonstrated a marginal increase in the AUC (from 0.77 to 0.78) in a model that included ABI in addition to classical risk factors, compared to risk factors alone; and a recent study [49] comprising about 500 consecutive asymptomatic patients without known atherosclerotic vascular disease, in whom ABI and IMT were measured, concluded that ABI measurements were not sensitive enough to be suitable for detecting subclinical atherosclerosis in these middle-aged individuals. Yet a recent meta-analysis showed that a low ABI (≤ 0.90) was associated with an approximately 2-fold increased risk to die from all-causes and cardiovascular causes as well as from any major coronary event in the various Framingham risk categories. Even more importantly, inclusion of the ABI in risk stratification in addition to the FRS would result in reclassification of the risk categories in approximately 19% of men and 36% of women [50].

In addition to the diagnostic tools discussed above, a number of other techniques have been suggested to improve risk prediction in various settings [51–53].

Comparative performance of measures of subclinical atherosclerosis

Of note, the degree of correlation between the various non-invasive measures of atherosclerosis in several vascular beds is modest at best. Simon et al. [54] carried out meta-analyses on the comparative performance of several subclinical atherosclerosis tests in predicting CHD in asymptomatic individuals, and they could clearly demonstrate that different types of tests convey different prognostic information. In addition, there are clear-cut differences in a number of important criteria for the selection of measurements of subclinical atherosclerosis, like predictive values, but also simplicity, reproducibility, safety and costs. Jacobs and Crow [55] also demonstrated only moderate correlation coefficients between coronary artery calcium scoring, IMT measurement, and ABI, thus strengthening the point that at the present time it is not clear which of the available methods should be preferably used for screening at least in subjects at intermediate risk. Such differences are likely to be due to variability of atherosclerosis manifestation rather than to be explained by poor measurements using either method [56]. In an attempt to overcome such caveats, Price et al. [57] combined ABI and IMT measurements and, although both methods showed similar accuracy in predicting CVD events, the combination of both further increased the AUC.

Combination of blood biomarkers and markers of subclinical disease

Since there is no consensus at present based on the available literature that either any of the before-mentioned blood biomarkers or the discussed measures of subclinical disease can be used for routine screening in unselected populations because of lack of supportive data, it is not surprising that for the combination of blood biomarkers and markers of subclinical disease, although basically an attractive concept, there is even less prospective data available. Indeed, only two studies have carried out such analyses, combining either CRP with coronary calcium scoring or with ultrasonographic IMT measurements. Park et al. [58] studied 1,461 participants without CHD at base-line for the presence of coronary calcium and measured CRP. During follow-up of 6.4 years, in subjects with base-line CRP ≤ 10 mg/L, the presence of coronary artery calcium was a predictor of cardiovascular outcome, as was CRP. Further analysis showed that there was an increasing risk with increasing calcium and CRP. Relative risks for the medium-calcium/low-CRP risk group to the high-calcium/high-CRP group ranged from 1.8 to 6.1 for MI/coronary death and from 2.8 to 7.1 for any cardiovascular event. Thus, the authors concluded that CRP and coronary calcium scoring both contributed independently towards the incidence of cardiovascular events. Most recently Cao et al. [59] presented important data from the CHS. The investigators simultaneously measured carotid IMT, plaque characteristics, and CRP and related all three variables to 12-year incidence of CVD events and all-cause mortality in 5,888 elderly subjects. The main results showed that all parameters were correlated with one another and each parameter independently predicted risk of CVD events and mortality in multivariable models which included all three measures and traditional risk factors. Being in the top tertile of the carotid IMT distribution was more predictive for various events than having a CRP > 3 mg/L or being in the high-risk group on the basis of carotid plaque characteristics. However, elevated CRP was a particularly useful predictor in the presence of subclinical atherosclerosis with a 72% increase in risk for CVD and a 52% increase in total mortality. Cumulative event rates suggested a possible additive interaction for incident CVD and all-cause mortality with an excess rate attributable to the interaction of CRP and subclinical disease of 54% for CVD death and 79% for all-cause mortality. By contrast, CRP did not add predictive power in the absence of carotid atherosclerosis. Finally, both CRP and subclinical atherosclerosis added only modest incremental information to risk prediction when adjusted for the effect of conventional risk factors with either C-statistics or AUC derived from ROC analysis. Thus, much more data of various blood biomarkers and non-invasive methods to assess subclinical disease are needed to further evaluate the potential of combining both for the improvement in cardiovascular risk prediction.

Limitations of the application of biomarkers at present

The rapidly increasing literature on biomarkers in cardiovascular disease has provided us with valuable new information regarding the pathophysiology of this complex disorder. Such information comes both from blood biomarkers as well as from markers of subclinical disease. However, before such information can be translated into the clinical setting, a number of criteria have to be fulfilled before any biomarker can be used routinely. It is not sufficient to simply demonstrate a more or less strong association between a biomarker and cardiovascular outcome [60]. Recently Morrow [61] has put together an extensive list of requirements that cover the pre-analytical era, assay methods, costs involved, strength of the association found in various studies, and the potential incremental value over and above existing parameters. In particular, detailed knowledge is needed regarding various test characteristics that should cover discrimination, calibration, and reclassification of subjects [62–64]. Unless such information has been convincingly shown in very large prospective studies in representative populations and randomized clinical trials, wide-spread application of biomarkers to refining risk predication cannot be recommended. Thus, the issue of whether or not blood biomarkers as well as markers of subclinical disease contribute incremental information above and beyond that gained by traditional clinical variables has not been unequivocally settled. In addition, for markers of subclinical disease as well as for blood biomarkers controversy exists: which parameter represents the most useful one, for which time period of the atherosclerotic process, and which combination of markers may be most appropriate. Among the many blood biomarkers under investigation at present, CRP, NT-proBNP, Lp-PLA₂, and cystatin C come closest to the criteria required for an acceptable clinical utility, but none so far has fulfilled all of them. Among markers of subclinical disease there is fairly strong evidence for coronary artery calcium (CAC) scoring that it might be useful for further risk stratification, but concerns relate to the potential risks from radiation exposure of the general population when wide-spread CAC screening is carried out.

Future perspectives

In the future, we will see new candidates discovered by proteomics [65] or other ‘omics’ approaches; we will deal with biomarker profiles that cover various aspects of the complex pathophysiology of the atherothrombotic disease using new multiplex technologies; we will focus more on biologic patterns or whole biological (sub)systems that may become dysbalanced during the atherosclerotic process. Functional molecular imaging may be able to overcome some of the present problems and may integrate various approaches using bioimaging and circulating molecular markers [66]. Together with such innovative biotechnical approaches we will also have to develop new analytical techniques to adequately analyse such biologic information using tools provided by systems biology [67]. Thus, there is still some way to go before we may have the ideal, reliable diagnostic tool in our hands that enables us to identify atherosclerosis at its earliest stage, with economically acceptable costs [68].

Acknowledgements

I would like to thank Dr. Natalie Khuseyinova for editorial assistance in preparing the manuscript. Declaration of interest: The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper.