Why, how and in whom should the clinician estimate total cardiovascular risk?

Introduction

Historically, cardiovascular (CV) risk factors have been managed in isolation. However, atherosclerotic cardiovascular diseases (CVD) are multifactorial in origin and risk factors interact. Therefore, prevention of CVD in a given person should be adapted to his or her total CV risk: the higher the risk, the more intense the action should be. That was the main reason for introducing the concept of “total CV risk” in European recommendations on the prevention of coronary heart disease in clinical practice in 1994 [1], analogous to those which had been proposed in New Zealand a year earlier [2]. The purpose was to stratify the population into different categories of total CV risk and to use total CV risk estimates in communications with the patient on CV risk management.

Nowadays, different international and national expert committees have adopted that strategy for the primary prevention of CVD, although there have been no randomized controlled trials of a total CV risk approach to either CV risk estimation or CV risk management.

Why risk stratification?

The reasons for risk stratification at the community level are multiple:

• To tailor the intensity of risk factor management to the severity of the total CV risk.
• To apply limited resources for preventive cardiology as cost-efficiently as possible.
• To consider cost-benefit or benefit-risk ratios in total CV risk management.
• To target a greater equity in the distribution of effective preventive actions.

There are other reasons to recommend the use of estimates of total CV risk for the prevention of CVD:

• For many years, it has been known that atherosclerosis and its clinical consequences are the result of multiple, interacting risk factors. These interactions are usually complex and not easy to assess clinically.
• Physicians treat patients, not isolated risk factors. The aim is to reduce the total CV risk of the patient, not only his elevated blood pressure or LDL cholesterol, and, if a goal cannot be reached with one risk factor, total CV risk can still be reduced by trying harder with other risk factors.

What is “total CV risk”?

The total CV risk of a given individual is his/her likelihood of developing a CV event over a certain time period. This chance is estimated on the basis of age and gender and a limited number of the major CV risk factors, and is generally expressed as a percent. Most of the models which are available to estimate total CV risk predict “hard” CV endpoints, fatal or non-fatal, and generally for time periods of 10 years or for life.

However, it should be emphasized that the CV risk estimation of a given person expressed in terms of absolute risk remains hazardous, very approximate and therefore, at present, elusive. The purpose of these models is to stratify the population into different CV risk categories and to apply preventive actions accordingly.

The problem is that clinicians and patients tend to look at these models on an individual basis while the strategies that are proposed are based on a group approach.

Another problem that is frequently encountered when total CV risk estimates are used occurs when total CV risk is handled as a dichotomy: those at high risk and all the others. This leads to an underuse of the total CV risk concept; total CV risk is a continuum. Not only those at high CV risk should receive attention; appropriate action is also needed in the majority of adults at moderate CV risk. It should be realized that all of the cut-offs which are used to define a low, moderate, high or very high total CV risk level are arbitrary and not evidence-based. They are based on practical considerations in relation to the healthcare system, health insurance plans and economic determinants. The choice of these cut-offs reflects the ability of the health system to care for persons at risk.

An elevated level could be defined as that level of total CV risk above which the chance of developing CVD is increased and above which a reduction of the total CV risk is more effective than harmful and also cost-efficient. Accordingly, one can speak of moderately elevated, highly elevated or very highly elevated total CV risk

How to estimate total CV risk?

There are subsets of patients who are at high or very high CV risk because of certain conditions. These patients will therefore need everything which we have available in order to help reduce their risk as much as possible. These subsets are the patients with documented atherosclerotic CVD, with type 2 diabetes, with very high levels of individual risk factors, such as familial hypercholesterolemia or severe arterial hypertension, and those with chronic kidney disease. They may represent 20-30% of the adult and elderly population. In these patients no risk estimation tools are needed.

Fortunately, the vast majority of the adult and elderly population is asymptomatic and apparently healthy. For these patients, one needs risk estimation models to stratify the population into those at low, moderate, high and very high total CV risk. Indeed, some of them may be at high risk because of a clustering of CV risk factors. However, those at moderate risk should also receive advice regarding lifestyle changes, and in some cases drug therapy will be needed to control arterial hypertension or dyslipidemias. In these subjects preventive actions are needed in order to:

• prevent further increases in total CV risk;
• increase awareness of the danger of CV risk;
• improve risk communication; and
• promote primary prevention efforts.

Low-risk individuals should be given advice to help them maintain their low risk status for as long as possible. Thus, the intensity of preventive actions should be tailored to the patient’s total CV risk.

To estimate total CV risk in an apparently healthy subject one should use good clinical judgment, but this alone is insufficient. It is critically important to use risk estimation tools because many people have mildly raised levels of several risk factors that, in combination and through multiplicative interaction, can result in unexpectedly high levels of total CV risk. In these people, decisions on drug treatment for arterial hypertension or elevated LDL cholesterol should not only be based on the level of the individual risk factor but on the total CV risk.

Which risk estimation model to use?

A very large number of CV risk scoring models have been developed. In a NICE guideline from 2007, the authors identified 110 different risk scoring methods in the literature up to 2005, of which 70 had been used in the primary prevention of CVD [3]. Since then even more have been developed. A few examples are listed here:

• The mother of all equations is the model developed in the Framingham study. Different equations have been developed [4] based on observations in more than 8,000 men and women in the Framingham Heart Study and the Framingham Offspring Study, including sex, age, total cholesterol (TC), HDL cholesterol (HDL-C), systolic blood pressure (SBP), smoking status, diabetes and antihypertensive drug therapy.  
• ASSIGN [5] is based on observations in more than 13,000 men and women from the general population in Scotland. It has the advantage of including in the model (besides sex, age, TC, HDL-C, SBP, smoking, and diabetes) an area-based index of deprivation and family history.
• QRISK1 [6] and QRISK2 [7] are based on large sets of data collected from GP databases in the UK. The 10-year risk of CVD events is estimated as well as the lifetime risk as is recommended in the JBS3 guidelines [8].
• SCORE [9] is a model based on 12 pooled prospective studies from 11 European countries. The 10-year risk of CVD mortality is estimated based on sex, age, TC, SBP and smoking status. Separate charts were developed including HDL-C. Versions for use in high- and low-risk countries in Europe are available as well as national, updated recalibrated versions.
• In parallel with the new guidelines on CVD prevention in 2013, the AHA/ACC developed a new risk estimation system, the pooled cohort studies equations [10]. Results are based on 4 pooled prospective studies from the USA with more than 25,000 men and women from different ethnic origins. The 10-year risk of a first atherosclerotic CV event is estimated as well as the lifetime risk based on sex, age, race, TC, HDL-C, SBP, antihypertensive therapy, diabetes and smoking.
• Recently, Globorisk was presented based on eight US cohorts [11]. It predicts fatal CVD but also fatal+non-fatal CVD. The equation can be recalibrated. The methodology helps to overcome technical barriers for global application of risk stratification.

The validity of some of these models has been tested. This frequently resulted in either underestimation or overestimation of total CV risk depending on the population in which they were tested [12]. From a review of the performance of different risk estimation models in the same population, it was concluded that the results were inconsistent and susceptible to potential biases and methodological shortcomings [13]. It is clear that one size does not fit all. The theoretical answer to the question “Which model to use in my clinical practice?” seems to be “the model that is based on recent observations in a population that resembles as closely as possible the community from which your patients are drawn”.

In daily practice, one has to realize that most models are based on observations in prospective studies that were carried out years ago. When applied to populations where the incidence of CVD has declined, they will overestimate CV risk; when used in populations where the disease is on the increase, total CV risk will be underestimated. In that respect, one should realize that the epidemic of CVD has been, and still is, very dynamic [14].

In whom should the clinician estimate total CV risk?

The estimation of total CV risk is recommended in all asymptomatic, apparently healthy individuals with an a priori higher CV risk, such as persons with a family history of premature CVD, in smokers of tobacco, those with elevated blood pressure, glucose intolerance, metabolic syndrome or dyslipidaemia or with comorbidity that places them at higher CV risk. In general, repeat CV risk estimation every five years is recommended. In persons with a total CV risk estimate close to a threshold that may implicate a change in management, repeated measurements may be needed more often. Opportunistic screening may be considered in men >40 years of age and in women >50 years of age (or post-menopausal at a younger age). Opportunistic screening in younger persons with no known CV risk factors is not recommended due to low cost-effectiveness.

Can these risk estimation models be improved by adding novel risk markers?

Besides the conventional risk factors which are included in most models, there is a long list of variables which are clearly related to the incidence of CVD. Some relate to lifestyle, and composites have been constructed on the basis of dietary factors. More objective measures are now available to quantify physical activity. Psychosocial factors related to stress or vital exhaustion are important, and other biomarkers have been proposed from the fields of lipid metabolism, inflammation or coagulation.

However, the demonstration of the predictive value of a factor, independent of the conventional CV risk factors, is, on its own, no proof of the incremental value of the marker in a risk estimation model. From the literature it can be concluded that the improvement in risk stratification and reclassification from adding novel markers separately or as multimarkers to models based on established CV risk factors is small and inconsistent [15,16].

Why does an independent strong risk factor not improve risk estimation?

Age and gender alone result in a high area under the receiver operating characteristic curve (AUROC) of up to 0.70. Estimation of risk is not a diagnostic test and will therefore never result in an AUROC of 1.0. The inclusion of the main risk factors - smoking, SBP and TC - brings the AUROC up to 0.80-0.85. This leaves little room for improvements in the AUROC by adding new risk markers.

Recent guidelines (17,18] have recommended considering other risk markers as qualifiers or risk modifiers. Some of these are listed in Table 1. The assessment of these factors can be useful if they improve risk classification and if they are easy to measure in daily practice. Particularly in individuals at intermediate or moderate total CV risk according to existing models, the variables listed in Table 1 may move an individual’s total CV risk upwards or downwards, and this may influence management decisions.

Table 1. Factors that may modify total CV risk estimates.

In considering all this, one should keep in mind:

• that all models assist in risk estimation but must be interpreted in the light of the clinician’s knowledge and experience; and
• that re-classification can be of value when the individual’s risk lies close to a risk threshold that may alter the preventive strategy.

There are also observations from cohort studies suggesting that total CV risk can be higher than indicated in the models in asymptomatic persons with markers of subclinical atherosclerotic vascular damage detected by coronary artery calcium (CAC), ankle-brachial index (ABI), pulse-wave velocity (PWV) or carotid ultrasonography. In studies comparing these markers, CAC had the best reclassification ability [19].

Therefore, the use of methods to detect these markers may be of interest in subjects who belong to the intermediate CV risk group. A reclassification into the low or high total CV risk categories may indeed influence the management strategy.

However, not all of these techniques have an excellent precision: radiation exposure may be harmful, and the cost-effectiveness in terms of better prevention has not been shown. If they are used, then cut-off values for considering these markers as modifiers of total CV risk are a CAC score of  >400 Agatston units, ABI of <0.9 or >1.40, aortic PWV of 10 m/s, and the presence of plaques at carotid ultrasonography.

CV risk estimation models can also be improved by recalibration. In Europe, where large differences in CVD incidence are present, recalibration at the national level can have great advantages. To do so, one needs recent and precise estimates of the risk factors and of the endpoints that are used in the model. The SCORE model has the advantage of estimating the risk of CVD mortality. In most European countries, reliable and recent mortality statistics are available for this purpose. Recalibrated country-specific versions of the SCORE model exist for Belgium, Bosnia and Herzegovina, Croatia, Cyprus, Czech Republic, Estonia, France, Germany, Greece, Poland, Romania, Russian Federation, Slovakia, Spain, Sweden and Turkey. In some countries the recalibrated models have been validated, and have performed, in general, better.

Other risk estimation systems can also be recalibrated, but the process is easier for mortality than for total events. The reasons for preferring a model that estimates fatal instead of total fatal+non-fatal events are that non-fatal events are dependent on definition, developments in diagnostic tests, and methods of ascertainment, all of which can vary over time and between studies, resulting in very variable multipliers to convert fatal to total events.

CV risk prediction models may also be used in clinical practice in communications with the patient on the most appropriate CV risk management strategy. From a systematic review of the effect of giving total coronary risk information to adults, the authors concluded that providing such information improves the accuracy of risk perception and probably increases the intent to initiate CVD prevention among adults at moderate to high risk [20].

These findings are consistent with behavioral theory and with reviews on the effect of health risk appraisal tools, suggesting that such information is rather a tool to be used in conjunction with other tools to promote adherence and risk reduction.

There is, however, a need to examine the best presentation of total CV risk to the patient. Insufficient attention has been given as to how one best to help individuals in making sense of risk information in the context of decision making. Particularly in young subjects who are at low absolute risk because of their age but who may already have several important risk factors, other measures of risk may be more useful. Take, for example, a young man from a low-risk country in Europe, 40 years old, a smoker, and who has an SBP of 160 mmHg and a TC of 7 mmol/L (280 mg/dL). According to the SCORE model, his total CV risk of dying from a CVD in the coming 10 years is only 1%. Will that figure be convincing to help him to stop smoking and to control his SBP and TC? More motivation could result from an estimate of his relative risk. His relative risk compared with a man of his age who does not smoke and has optimal SBP and TC levels is six times greater. That figure may be more convincing in motivating him to make the necessary lifestyle changes.

Another approach to this problem is to use heart age, also called risk age or cardiovascular age. This is the age of a person with a given burden of risk factors compared with the age of a person who has fewer risk factors. Based on the SCORE chart for high-risk countries in Europe, the risk of fatal CVD for Mr X, a 40-year-old male smoker with an SBP of 180 mmHg and a TC of 7 mmol/L (280 mg/dL), is 3%/10 years, which is similar to that of a 60-year-old non-smoking man with an SBP of 120 mmHg and a TC of 4 mmol/L (160 mg/dL); therefore, the heart age of Mr X is 60 years.

Risk age is an easily understood way of illustrating the likely reduction in life expectancy that a young person with a low absolute but high relative risk of CVD will be exposed to if preventive measures are not adopted. Risk age has been shown to be independent of the CV endpoint used, which bypasses the dilemma of whether to use a risk estimation model based on fatal CVD only or the less reliable endpoint of total CVD events. Risk age can be used in any population regardless of baseline risk and of secular changes in mortality, and therefore avoids the need for recalibration.

Another useful tool in communicating risk to individuals with high risk factor levels but a low 10-year absolute risk of CV events is an estimate of lifetime risk [21]. The number of risk factors and their interaction continue to play a role, but lifetime risk will be higher in younger persons because of their longer exposure times. However, the role of lifetime risk in treatment decisions has not yet been established; therefore, it should at this moment only be used in communicating risk estimations and management.

Conclusion

In conclusion, most risk estimation models to estimate total CV risk perform similarly when applied to populations recognizably comparable to those from which the risk estimation system was derived. The challenge is not the need for a more personalized prevention but the failure to act in those who have the potential to benefit. The existing models should be used for the purpose for which they were developed. By adding other risk markers, performance may not improve greatly, but this may be useful in re-classification of subjects at intermediate risk. Finally, total CV risk should be handled as a continuum not by dichotomizing into high versus the rest.