The interplay between risk factors, the variable influence of any one risk factor and the tendency for risk factors to cluster make the determination of cardiovascular disease (CVD) risk in apparently healthy individuals more complex1 than assessing any single risk factor in isolation. For example, age factors heavily in a patient’s CVD risk; however, a younger patient with multiple factors may also be at high lifetime CVD risk. Calculating that a younger patient has low 10-year CVD risk may cause the clinician to miss an opportunity to address modifiable risk factors prior to the onset of disease. Risk-factor clustering may also lead to an underestimation of CVD risk. One well-known, though controversial, example of risk-factor clustering is the metabolic syndrome. Metabolic syndrome increases the risk of CVD mortality in middle-aged men by as much as three-fold.2 If the full spectrum of metabolic syndrome components – i.e. central obesity, hypertension, low high-density lipoprotein cholesterol (HDL-C), elevated triglycerides and increased blood sugar – were not assessed, the clinician would fail to recognise the severity of the patient’s CVD risk.
The concept of assessing multiple risk factors concurrently is referred to as total CVD risk, expressed as absolute risk within a defined time period. Total CVD risk assessment is useful in assisting clinicians in the prioritisation of treatment in patients who would derive the greatest benefit (i.e. high-risk patients), while reducing unnecessary drug exposure in low-risk patients. In addition, risk estimation provides a means of displaying total CVD risk in consultations between clinicians and patients in which lifestyle changes or other therapeutic interventions are recommended. The identification of coronary heart disease (CHD) and CHD risk equivalents can further simplify the CVD treatment algorithm for time-constrained practitioners and target intensive therapy appropriately in secondary prevention patients. The recent presentation comparing the results from the European Action on Secondary Prevention through Intervention to Reduce Events (EUROASPIRE) surveys highlights a failure to address CVD risk factors such as hyperlipidaemia, hypertension, obesity, diabetes and smoking cessation in clinical practice.3 These findings were in the context of the more clearly defined area of secondary prevention; the challenges are even greater in primary prevention.
In an effort to simplify the process of total CVD risk assessment, a number of risk-assessment tools have been developed, several of which are available as online calculators or risk charts (see Table 1). Each risk-assessment tool calculates risk according to a distinct algorithm, and may be used preferentially in various countries or when evaluating specific patient populations. The recent introduction of the new European guidelines on CVD prevention in clinical practice provides an ideal starting point from which to discuss total CVD risk and the similarities and differences in methodology employed for risk assessment throughout Europe.
Methods of Risk Assessment – Comparisons Among Guidelines
The first step in evaluating CVD risk is to identify patients who are candidates for total CVD risk assessment. In the new European guidelines, certain patients – such as those with established atherosclerotic disease, diabetes or marked increases in a single risk factor and end-organ damage – are automatically classified as ‘high-risk’ and require full preventative intervention.4 These patients were traditionally considered to be eligible for secondary prevention. However, in individuals who do not meet these criteria the degree of CVD risk is less clear-cut. These individuals may also be at high risk of CVD because of an accumulation of multiple risk factors rather than the presence of a single high-risk attribute.
The categories of patients who are recommended for total CVD risk assessment vary according to the CVD prevention guidelines used. The European guidelines on CVD prevention have established a hierarchy for prioritising CVD prevention in clinical practice.4 Patients identified a priori as high-risk because of the presence of one or more of the above-mentioned risk factors are automatic candidates for therapeutic intervention. In patients without these risk factors, total CVD risk assessment is recommended if requested by the patient, in patients who are middle-aged smokers, in patients with one or more known risk factors, in patients with a positive family history and in patients with symptoms suggestive of CVD.
The Joint British Societies’ (JBS) guidelines on the prevention of CVD are based on clinical best practice and are more broad in scope, recommending risk assessment starting at the age of 40 years for individuals without a history of CVD or diabetes.5 Therapeutic intervention is then advised in those with a 10-year CVD risk of ≥20%. According to the JBS guidelines, individuals with hypertension, a total cholesterol to HDL-C ratio of ≥6 and those with familial hyperlipidaemia also require therapeutic intervention. However, interpreting the authors’ intentions regarding treatment of the identified risk factor versus treatment of all risk factors is not straightforward because the guidelines state that risk scoring should be performed on pre-treatment levels of blood pressure and cholesterol for those already receiving medication.
Once a clinician has identified a patient for total CVD risk scoring, there are many risk-scoring options available, some of which may be preferentially suggested by CVD prevention guidelines or country-specific cardiology societies (see Table 2).4–9 Risk scoring can be as simple as totalling the number of risk factors, or can be complex, requiring the use of formulae that account for the differences in relative contribution of the various risk factors.
The Systematic Coronary Risk Evaluation (SCORE) risk charts are the most widely used of the risk-scoring tools, as they are recommended by the European Society of Cardiology (ESC) and cardiology societies from Austria, Belgium, Finland, Germany, The Netherlands, Portugal, Russia, Spain, Sweden and Switzerland. The SCORE project was commissioned by the ESC and the Second Joint Task Force, and was first adopted in the 2003 edition of the ESC guidelines on the prevention of CVD.10 The SCORE algorithm was derived from a pooled data set of 205,178 individuals from 12 European cohort studies that represented 11 countries, giving it the advantage of including a broad range of patient populations but only data on CV mortality.11
SCORE charts provide an estimation of the 10-year risk of fatal CVD based on an individual’s age, systolic blood pressure, total cholesterol or total cholesterol to HDL-C ratio, gender and smoking status (see Table 3), and include separate risk-scoring charts for high-risk and low-risk countries. This designation is based on cardiovascular death rates and age-standardised death rates in national mortality statistics. Even with this adjustment for regional variability, further calibration of the SCORE charts was required for several countries, including Belgium, Germany, Greece, The Netherlands, Poland, Spain and Sweden; this is reflected in the recent version of the ESC guidelines.4
Each of the risk-scoring algorithms listed in Table 3 accounts for age, blood pressure and/or hypertension, cholesterol and smoking status when determining an individual’s risk.5,6,11–16 The particular cholesterol measure, however, is highly variable, with different algorithms including total cholesterol, low-density lipoprotein cholesterol (LDL-C), HDL-C or some combination thereof. The family history parameter is also highly variable and algorithm-dependent. Family history may include premature CHD, CVD, myocardial infarction (MI) and/or sudden death. The Prospective Cardiovascular Munster Study (PROCAM) is the only risk-scoring tool that does not consider gender. This is because the original study on which it is based enrolled only male participants.15
The Scottish Intercollegiate Guidelines Network (SIGN) developed its own risk-scoring tool – Assessing cardiovascular risk using SIGN guidelines (ASSIGN) – to address two areas it felt were not covered in conventional risk-scoring algorithms: the Framingham score for social deprivation and family history.12 Although family history is considered by several of the risk-scoring algorithms, social deprivation is a relatively novel risk factor (see Table 3).
Social deprivation refers to low-income populations, who may have limited access to healthcare and health education. Previous studies have demonstrated that the Framingham score underestimates CVD risk in socioeconomically deprived individuals.17,18 In ASSIGN, social deprivation is estimated by geographical region using the Scottish Index of Multiple Deprivation (SIMD).12 Compared with Framingham, ASSIGN classified more socially deprived individuals as high-risk based on 10-year CVD risk.
QRISK is a new risk-scoring algorithm designed to improve the prognostic value of risk scoring among patients in the UK and account for the impact of social deprivation on CVD risk.16 The QRISK scoring algorithm was developed from a cohort of 1.28 million individuals in the UK who were between the ages of 35 and 74 years and who were free of diabetes or CVD at the time of enrolment. Data on the first diagnosis of CVD, including MI, CHD, stroke or transient ischaemic attack, were available for 8.2 million person-years of observation. A cohort of 610,000 UK individuals from an age- and gender-balanced population was used to validate the QRISK risk-scoring algorithm. QRISK shares many parameters with other risk-scoring algorithms, but also includes the Townsend score, which, like the SIMD, is a surrogate measure of social deprivation based on geographical region (see Table 3). The threshold for high risk with QRISK is a 20% or greater chance of developing CVD within the next 10 years. Compared with the Framingham and ASSIGN risk-scoring tools, QRISK more accurately assigned risk level in this UK population; in contrast, both Framingham and ASSIGN were predisposed to overestimating CVD risk.
Another differentiating point among risk-scoring algorithms is the type of risk predicted. Framingham, the National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) (NCEP ATP III) and PROCAM all evaluate 10-year risk of a coronary event.5,14,19 ASSIGN, the Italian Risk Charts, JBS 2, QRISK and SCORE are more broad in that they assess 10-year risk of CVD, which is inclusive of both coronary and cerebrovascular events (e.g. stroke or transient cerebral ischaemia).5,11–13,16 It is notable that SCORE is predictive of fatal events, whereas all of the other algorithms mentioned are predictive of both fatal and non-fatal events.11 In addition, SCORE does not provide a threshold for high risk; rather, data in the SCORE risk charts are provided as percentages and colour-coded groupings.
The Italian risk charts, which are derived from the CUORE Project, present risk levels in a slightly different form as well, with six colour-coded levels of risk rather than distinct numerical cut-offs. This is driven by the variability in risk ranges per grouping for men and women.13 Because not all risk-scoring algorithms include all potential risk factors, CVD prevention guidelines acknowledge that the clinical judgement of the physician, which may include consideration of additional risk factors, is imperative.
The ESC guidelines add qualifiers for risk estimation using SCORE by stating that sedentary or obese individuals, those with a strong family history of CVD, those who are socially deprived and those who have diabetes, low HDL-C or high triglycerides or evidence of pre-clinical atherosclerosis will be at higher risk than indicated in the SCORE risk chart.4 Similarly, the JBS 2 and NCEP ATP III guidelines recognise patient characteristics such as ethnicity, obesity, waist circumference, physical inactivity, atherogenic diet, lipoprotein(a), homocysteine, pro-thrombotic and pro-inflammatory factors, impaired fasting glucose and evidence of subclinical atherosclerotic disease as additional CVD risk factors that are not directly addressed in the risk-scoring algorithm, but that should be considered when evaluating an individual patient’s risk.5,14
Risk Scoring in Special Populations
Risk-scoring tools are developed using data from specific, and often limited, patient cohorts. The Framingham score, which is the basis for 10-year CVD risk assessment recommended in the JBS 2, the National Institute for Health and Clinical Excellence (NICE) and NCEP ATP III guidelines,5,9,14 is derived from 5,573 men and women aged 30–74 years who participated in the Framingham Heart Study.19 One of the criticisms of the Framingham algorithm is that it was based on a predominantly white, US-based cohort, which may not accurately reflect the CVD risk factors of greatest importance in a European population. In addition, the Framingham score was developed from data gathered during the peak of CVD in the US.
The PROCAM algorithm, which is used predominantly in Germany and Austria, is also based on a limited population: 5,389 men aged 35–65 years who participated in the original PROCAM study.15 These limitations call into question the broad applicability of a risk-scoring algorithm outside the population in which it was developed. Moreover, application of a single tool may not provide accurate risk prediction across all populations and ethnic groups.
Several investigators have addressed the need to consider the influence of ethnicity on CVD risk by creating modified versions of existing risk-scoring algorithms.20–22 Risk-scoring algorithms have also been reported to over- or underestimate CVD risk in certain countries.23–26 The reasons for failure to predict risk adequately in these populations may relate to particular regional risks (e.g. high incidence of stroke in Portugal) or the failure to consider cultural, social, behavioural or genetic differences between populations. Modifications of risk calculators have been undertaken to account for regional differences in risk.
HellenicSCORE used local data on mortality and risk-factor prevalence to calibrate the SCORE algorithm for individuals in Greece.27 As previously discussed, SCORE calibrations have also been generated for Belgium, Germany, The Netherlands, Poland, Spain and Sweden.4 Another alternative to calibrating a widely used risk-scoring tool is the creation of a risk-scoring algorithm using data from the population in which it will be used, as is the case for ASSIGN, the Italian risk charts, QRISK and PROCAM.
Consistency Among Risk-prediction Models
As would be expected, risk-scoring algorithms share many CVD risk factors (see Table 3). Moreover, different risk-scoring tools may be based on data from the same epidemiological studies – for example, those that are based on the Framingham study. However, risk-score results, and hence treatment recommendations, vary by calculation method, even within the same population.28–30 In a head-to-head comparison of different guideline-recommended risk-scoring algorithms in a cohort of 100 consecutive patients without CVD who were referred to an outpatient lipid and diabetes clinic in Germany, the percentage of patients classified as high-risk varied from 26 to 53% depending on the algorithm used.28 Even more divergent was the percentage of patients in whom lipid-lowering therapy was recommended, ranging from 5 to 52%. Notably, versions of the ESC and JBS guidelines used in this study did not reflect the current iterations. It remains to be seen how the newer calibrated or specialised risk-scoring algorithms would perform in a ‘real world’ setting.
The consistent finding recognised by all of the CVD prevention guidelines is that CVD risk increases with the number of risk factors present. Risk-scoring algorithms provide an easy method for assessing major CVD risk factors in totality. As risk is a continuum it is difficult to create definite thresholds for establishing when therapeutic intervention should be considered for an individual patient.
A patient at increased risk but who does not meet the criteria for drug therapies may benefit from lifestyle changes to reduce risk-factor burden, and the clinician may want to follow this patient more closely, performing routine CVD risk assessments. Risk scoring provides a guide, but the decision to treat ultimately resides with the clinician.
Currently adopted risk scores, with the focus on 10-year risk, are dominated by the impact of age, giving a disadvantage to women and younger individuals at high lifetime CVD risk. An important opportunity exists to revise risk prediction, taking lifetime risk into account. The issue is critically important as a stimulus to risk-factor and lifestyle modification in younger individuals at high risk of developing CVD. For example, a 40-year-old woman with hypertension, diabetes and a smoking habit would not qualify for prevention therapy because her 10-year risk is very low (well below 20%), despite the fact that her lifetime risk is very high. The current emphasis on 10-year risk has the impact of delaying risk-factor modification until older age, when many of the changes are more difficult to address. To account for this disparity, lifetime risk estimation has been recommended as an adjunct to the standard 10-year risk determination.31 The proposal is that every patient should have not only their 10-year risk but also their lifetime risk (expressed as life-years lost) assessed.
The decision to treat is usually based on high absolute risk. High relative risk is something that a physician should use to encourage the patient to change his or her lifestyle and to comply with therapeutic interventions. The ESC guidelines provide a relative risk chart for SCORE to demonstrate the effect of CVD risk factors on total CVD risk independent of age, using the average risk for that age as a comparator.4
Another approach to addressing the impact of age is to project risk for a younger patient to his or her risk at a more advanced age. The JBS 2 guidelines recommend projecting risk in individuals aged <50 years to their risk at the age of 49 years.5 Whatever the means, it is important to communicate to a younger individual with multiple CVD risk factors his or her long-term prognosis – this may have a powerful impact on encouraging lifestyle and risk-factor modification.
Cost-effectiveness of Risk Scoring
The cost-effectiveness of risk scoring is not typically addressed by CVD prevention guidelines. There is concern that risk scoring will substantially increase the number of patients who are eligible for drug therapy,32 thereby increasing healthcare costs. However, studies have shown that both pharmacological and non-pharmacological interventions for the prevention of CVD are cost-effective, particularly in high-risk individuals.33–36 Lifestyle modifications, such as smoking cessation, are the most cost-effective of the therapeutic interventions available to patients at elevated CVD risk.35 Much of the benefit gained from CVD prevention is measured in terms of gains in life expectancy or quality-adjusted life-years. Reducing the risk factors that contribute to CVD risk has been shown to extend life expectancy.33–36
Despite the differences among risk-scoring algorithms and the limitations inherent within each risk-scoring tool, CVD risk-score assessment is a useful guide for determining patient risk and prioritising the need for appropriate lifestyle and/or medical interventions.
To address needs specific to a particular country, region or ethnic group, global risk-scoring algorithms may need to be calibrated within individual patient populations. Some of these modifications have already been created in terms of population-specific risk charts.
Regardless of the risk-scoring method chosen, it is important that clinicians increase their use of such systematic methods of prioritising which patients have the most to gain from intervention. However, until more accurate predictors at the patient level become available, clinical judgement should interpret risk-score findings in the clinical context of the patient.