Cardiovascular Risk Assessment: Making Sense of an Ever-Expanding Universe

In 2013, an estimated 600,000 cases of deaths in the United States were attributed to cardiovascular disease (CVD).1 With obesity, hypertension, and dyslipidemia rising in prevalence globally, improvements in early detection and treatment of CVD are strongly desired. In addition to better diagnostics and therapeutics, prevention is critically important to reducing the public health burden of CVD in the 21st century. The ability to prevent disease requires widespread public dissemination of healthy lifestyle habits and avoidance of known deleterious behaviors, the so-called "sick population" approach.2 A risk-based targeted approach to primary prevention, in which clinicians tailor the aggressiveness of interventions to each patient's risk for cardiovascular disease, is also recommended to optimize health.

A prerequisite for a personalized framework to primary prevention is an accurate prediction algorithm that can define those at highest risk of future disease, including individuals with advanced subclinical disease. Numerous risk prediction formulas have been developed to quantify the 10-year risk of CVD, including the 2008 Framingham Risk Score for global CVD events and the 2013 Pooled Cohorts Equation for atherosclerotic CVD (ASCVD). These formulas, widely recommended in clinical practice guidelines, can help inform decisions pertaining to the prevention of cardiovascular disease (i.e., use of aspirin and lipid-lowering agents). Yet, their predictive abilities to discriminate between individuals who later develop ASCVD and those who do not is estimated as only fair to moderate.3 Traditional risk assessment tools have also been shown to overestimate the risk of future CVD in both men and women from contemporary populations.3 In addition, nearly 15% of patients with myocardial infarction have no known traditional risk factors for coronary heart disease prior to their event and may have alternative processes contributing to their cardiovascular risk.4

One strategy to enhance the discriminatory ability of risk prediction equations is the use of circulating biomarkers. Serological assays of proteins that are central in the pathobiology of atherosclerosis have been shown in some studies to have relatively high predictive value for CVD. Indeed, the number of protein-based biochemical assays demonstrated to have strong associations with heart disease continues to grow at a fast pace and along with this come questions of their practicality and ease of use. Lipid fractions are the most notable contemporary example of blood-based biomarkers with a potent predictive potential. In the assessment of venous thromboembolism, fibrin-degradable products such as D-Dimer are routinely used to quantify the extent of endogenous fibrinolysis and thus the likelihood of thrombosis. As markers of inflammation, erythrocyte sedimentation rate and C-reactive protein are utilized, although correlation with heart and vascular tissue damage has limited specificity. For ischemia, D-lactate is an effective biological assay in the intestinal territory, while high-sensitivity troponin is used to gauge the severity of myocardial under-perfusion. New markers continue to emerge in all domains of cardiovascular medicine: serum fibrinogen levels have been shown to carry a strong predictive potential for stroke and cardiovascular disease.5 Plasma cystatin C is associated with higher rates of renal failure.5 Galectin-3 is predictive of heart failure.

Meaningful data about the likelihood of future CVD may also be derived from imaging technologies. Coronary artery calcification (CAC), a sign of subclinical coronary atherosclerosis, can be discerned decades before the acute clinical manifestations of coronary ischemia. The timely recognition of CAC, followed by the evaluation of its extent and anatomical distribution, allows for introduction of lifestyle changes or the initiation of lipid-lowering treatment for high-risk individuals.6-8

In addition to quantification of CAC using non-contrast cardiac CT, molecular imaging is now emerging as a promising avenue towards more granular functional-physiological cardiovascular assessment. This form of imaging entails the infusion of a sensitive and specific agent to the target of choice followed by the analysis of its distribution and spatial resolution. Recently described non-invasive targeted modalities of that kind could shed light on biological processes that are fundamental to normal and abnormal cardiovascular function. To take just a few novel examples: nuclear imaging of 99mTc-labeled annexin-V provides excellent correlations with percent macrophages and overall plaque stability of carotid plaques in mice;9 isotope-cell tracking can simulate the degree of monocyte accumulation in mice with atherosclerosis using SPECT and CT imaging;10 and collagen-targeted MRI contrast media can highlight fibrotic areas, delineating post-infarction myocardial scarring.11 Therefore, the advent of molecular imaging could greatly enhance our future cardiovascular risk prediction capacity in ways previously unimaginable.9

Genetic risk assessment is yet another strategy under study as an ASCVD risk assessment adjunct. For example, a multilocus genetic risk score (MGRS) obtained from genome-wide explorations of variants indicative of future CVD risk offers what is hoped to be a biologically-precise means to personalize clinical risk. Studies have shown that this MGRS is associated with coronary heart disease risk (HR ~1.53; P<0.001) and enables better risk classification when compared to conventional risk factors.12 However, in other single-nucleotide investigations, the genetic risk score failed to improve the predictive ability of traditional risk factors.13 As such, the ability to distinguish between clinically-important genomic variations and other nucleotide changes that are unlikely to have clinical ramifications must be acquired in order for genetic risk scores to provide additional refinement of risk prediction. Nonetheless, the promise of individual risk stratification that is based on genetic comprehension certainly exists.

A Multimodality Strategy to Assess Cardiovascular Risk – Better Risk Prediction?

Combining testing modalities that assess various non-redundant domains of cardiovascular pathology in order to generate a single risk score could greatly augment the sensitivity of traditional CVD assessment models. A recently published study by de Lemos et al. reveals the potential feasibility of such a risk assessment strategy using a hierarchy of cardiovascular risk factors drawn from a spectrum of testing strategies.14 The study's investigators selected five parameters thought to reflect distinct pathological processes intrinsic to cardiovascular disease: EKG evidence of left ventricular hypertrophy, CAC, N-terminal pro B-type natriuretic peptide (NT-proBNP), high-sensitivity cardiac troponin T (hs-cTnT), and high-sensitivity C-reactive protein (hs-CRP).

Participants from the Multi-Ethnic Study of Atherosclerosis (MESA; n=6621) and the Dallas Heart Study (DHS; n=2202) who were free of any CVD were followed for over 10 years, beginning in 2000-2002. The median age of participants at enrolment was 62 (MESA) and 44 (DHS) years. During the study period, 1,026 global CVD events occurred in the MESA and 179 in the DHS cohort. The investigators found that each of the five tests were associated with the primary global CVD outcome (a composite of cardiovascular death, MI, stroke, coronary or peripheral revascularization, incident heart failure, or atrial fibrillation) after risk factor adjustment, with the exception of hsCRP which did not correlate with the global CVD outcome in the DHS group.

For the model that included the five additional parameters, improvements in Harrell's C-statistic were seen for each cardiovascular endpoint (i.e., C-statistic for the global CVD outcome increased from 0.743 to 0.786). All of the five tests except hsCRP were also independently associated with the secondary composite ASCVD endpoint in MESA. CAC was the best predictor of coronary heart disease in both registries, followed by NTproBNP and hs-cTnT. Of note, strengths of associations were significantly attenuated in the case of hsCRP and hsTnT in the DHS cohort. All five tests were independently associated with incident heart failure, and the largest hazard ratios were seen for NTproBNP, EKG-LVH, and hscTnT. In addition, an integer score (from 0-5) that reflects the number of abnormal test results showed graded associations between increasing scores and the global CVD and ASCVD endpoints across sexes and ethnic groups.


In the study by de Lemos et al., the addition of five risk assessment tests to the traditional risk score enhanced cardiovascular risk prediction over traditional risk factors alone among adults free of known CVD enrolled in two prospective cohort studies. However, it is important to note that the base model which the investigators used was comprised of traditional risk factors (including age, sex, race/ethnicity, smoking status, diabetes, total cholesterol, HDL-cholesterol, systolic blood pressure, blood pressure medications, and statin use) and was specific to the sample under investigation. The increased predictive capacity of this multimodality score approach in routine care was not fully investigated, as the authors simply compared this score to traditional risk factors entered into a model applied to the sample under study rather than also comparing how the multimodality score compared to the pooled cohort equation.

Stepping back, do we need yet another score, or should we direct efforts instead to implementation of current best risk scores and tools in practice? Why derive yet another score when, for example, CAC, which is the strongest nontraditional risk factor, has yet to be subjected to an adequately powered randomised controlled trial that evaluates the benefit of adding CAC information to preventive management decisions?7

For the five parameter multimodality combination to be of incremental clinical merit, its discriminatory capacity must be compared to that of the standard prediction equations currently endorsed by the American College of Cardiology and the American Heart Association. In addition, some of the test results were significantly less predictive than others, and hsCRP had weak or no associations with numerous CVD outcomes in the both cohorts. The contributions of the individual tests for the different CVD outcomes varied, which further raises questions about the appropriateness of such a heterogeneous multimodality strategy to predict future CVD events, which are themselves heterogeneous in nature (myocardial infarction is quite different than heart failure, for example).

A formula that combines the use of various parameters, each with distinct correlations to different component outcomes of the CVD composite, may not facilitate the effective targetting of particular disease processes. For instance, an individual with two abnormal results--such as LVH and hsCRP--would likely not have the same global CVD outcome as that of a participants with strongly abnormal CAC and NT-proBNP. Similarly, the likelihood of coronary heart disease in an individual with significantly elevated hs-cTnT would probably be higher than that of one with only a mild elevation of hsCRP. By integrating cardiovascular test results from across the board and attempting to reflect upon a composite of cardiovascular end-points, the framework introduces room for false positive assessment while compounding the true risk of individual components of CVD outcomes.

It is also worthwhile to reflect on the broader clinical-scientific landscape in which we now practice. The current explosion of data in cardiovascular medicine is obvious. The number of currently-available parameters in association with an individual's risk of future ASCVD and CVD has reached overwhelming proportions.15 There is already a profound overabundance of independent predictive tools, which will only continue to grow with the evolution of our understanding of CVD. As the protein-based assays, nuclear-imaging patterns, and genetic markers described briefly above continue to proliferate, new models that incorporate such factors will become even more cumbersome to use and challenging to interpret for the practicing clinician.

There are probably thousands of genes relevant to each pathological process in the heart, as well as hundreds of circulating biomarkers and dozens of radiographic parameters whose addition to current algorithms could improve delineation of risk profile. But will theoretically granular multimodality prediction formulas be cost effective if they require multiple expensive testing modalities to function? Would it be practical for clinical use or rather serve a pure research purpose? And, more broadly, how can we ensure simplicity in care in which we effectively funnel our clinical expertise and technological insights to the true needs of the patient? Looking at the other side of the coin, how can we avoid dangerous convolution of care; a technologically-laden, patient-at-a-distance evaluation that risks treating the disease rather than the patient?

The answer, we believe, lies in recognition of high-yield CVD risk factors, those with consistent and definite linear relationships to biology, physiology, and pathology. CAC, for instance, reflects the atherosclerotic burden in the coronary arteries, adequately re-classifies at-risk patients compared with traditional risk factors alone, and is linearly associated with greater risks of a myriad of cardiovascular diseases. The inclusion of CAC in the MESA risk score was shown to improve risk prediction (C-statistic increasing from 0.75 to 0.8) with very good discrimination and calibration.16 Studies that will further test the concept of a CAC-based risk assessment and evaluate its comparative clinical performance in relation to that of traditional equations will permit more ideal utilization of the CAC tool and are therefore urgently needed. Inversely, the reliance on anecdotal parameters without established link to disease biology should be discouraged. The road to a simple, all-encompassing tool that would predict risk and guide preventive efforts is long and challenging.


  1. Centers for Disease Control and Prevention. Heart Disease Fact Sheet. Accessed April, 13, 2017
  2. Rose G. Sick individuals and sick populations. Int J Epidemiol 2001;30:427-32.
  3. DeFilippis AP, Young R, Carrubba CJ, et al. An analysis of calibration and discrimination among multiple cardiovascular risk scores in a modern multiethnic cohort. Ann Intern Med 2015;162:266-75.
  4. Yusuf S, Hawken S, Ounpuu S, et al. Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study. Lancet 2004;364:937-52.
  5. Van Holten TC, Waanders LF, de Groot PG, et al. Circulating biomarkers for predicting cardiovascular disease risk; a systematic review and comprehensive overview of meta-analyses. PLoS One 2013;8:e62080.
  6. Rozanski A, Gransar H, Shaw LJ, et al. Impact of coronary artery calcium scanning on coronary risk factors and downstream testing the EISNER (Early Identification of Subclinical Atherosclerosis by Noninvasive Imaging Research) prospective randomized trial. J Am Coll Cardiol 2011;57:1622-32.
  7. McEvoy JW, Martin SS, Blaha MJ, et al. The case for and against a coronary artery calcium trial: means, motive, and opportunity. JACC Cardiovasc Imaging 2016;9:994-1002.
  8. McEvoy JW, Diamond GA, Detrano RC, et al. Risks and the physics of clinical prediction. Am J Cardiol 2014;113:1429-35.
  9. Tawakol A, Osborne MT, Fayad ZA. Molecular imaging of atheroma: deciphering how and when to use 18f-sodium fluoride and 18f-fluorodeoxyglucose. Circ Cardiovasc Imaging 2017;10:e006183.
  10. Tekebe Y, Luma J, Einstein AJ, et al. A novel monoclonal antibody for RAGE-directed imaging identifies accelerated atherosclerosis in diabetes. J Nucl Med 2010;51:92-7.
  11. Caravan P, Das B, Dumas S, et al. Collagen-targeted MRI contrast agent for molecular imaging of fibrosis. Angew Chem Int Ed Engl 2007;46:8171-3.
  12. Ripatti S, Tikkanen E, Orho-Melander M, et al. A multilocus genetic risk score for coronary heart disease: case-control and prospective cohort analyses. Lancet 2010;376:1393-400.
  13. Damani S. Genetic risk scores for heart eisease: useful or useless? Sci Transl Med 2010;1 :EC29.
  14. de Lemos JA, Ayers CR, Levine BD, et al. A multimodality strategy for cardiovascular risk assessment: performance in two population-based cohorts. Circulation 2017. [Epub ahead of print]
  15. Fuster V. The proliferation of scoring systems. J Am Coll Cardiol 2017;69:1640-1.
  16. McClelland RL, Jorgensen NW, Budoff M, et al. 10-Year coronary heart disease risk prediction using coronary artery calcium and traditional risk factors: derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) with validation in the HNR (Heinz Nixdorf Recall) study and the DHS (Dallas Heart Study). J Am Coll Cardiol 2015;66:1643-53.

Clinical Topics: Anticoagulation Management, Arrhythmias and Clinical EP, Diabetes and Cardiometabolic Disease, Dyslipidemia, Heart Failure and Cardiomyopathies, Noninvasive Imaging, Prevention, Pulmonary Hypertension and Venous Thromboembolism, Anticoagulation Management and Atrial Fibrillation, Anticoagulation Management and Venothromboembolism, Atrial Fibrillation/Supraventricular Arrhythmias, Lipid Metabolism, Nonstatins, Novel Agents, Statins, Acute Heart Failure, Heart Failure and Cardiac Biomarkers, Computed Tomography, Nuclear Imaging, Hypertension, Smoking

Keywords: Atherosclerosis, Atrial Fibrillation, Biological Assay, Blood Pressure, Blood Sedimentation, Cardiovascular Diseases, Cholesterol, HDL, Coronary Artery Disease, Diabetes Mellitus, Dyslipidemias, Electrocardiography, Fibrin Fibrinogen Degradation Products, Fibrinolysis, Heart Failure, Hydroxymethylglutaryl-CoA Reductase Inhibitors, Hypertension, Hypertrophy, Left Ventricular, Myocardial Infarction, Infarction, Life Style, Natriuretic Peptide, Brain, Obesity, Peptide Fragments, Primary Prevention, Renal Insufficiency, Risk Assessment, Risk Factors, Smoking, Stroke, Thrombosis, Tomography, Emission-Computed, Single-Photon, Tomography, X-Ray Computed, Troponin I, Troponin T, Venous Thromboembolism

< Back to Listings