OUP user menu

Calculating global risk: the key to intervention

Gerd Assmann
DOI: http://dx.doi.org/10.1093/eurheartj/sui037 F9-F14 First published online: 30 June 2005

Abstract

The assessment of overall cardiovascular risk is a valuable and accepted means of identifying patients who are likely to benefit most from intervention to control individual cardiovascular risk. A straightforward and accurate means of calculating overall cardiovascular risk has been derived from the PROspective CArdiovascular Münster (PROCAM) Study, a large observational cohort study. The PROCAM risk calculator has several advantages over other risk calculators. For example, the algorithm derived from the Framingham study consistently and markedly overestimates the risk of myocardial infarction in the PROCAM population, or in a German cohort within the Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) Study, administered by the World Health Organization. In addition, the PROCAM risk calculator incorporates a broader range of commonly measured diagnostic parameters than the Framingham risk calculator. The multivariate Cox proportional hazards model that supports the PROCAM risk calculator has also provided valuable information on the principal driving forces behind atherosclerotic cardiovascular disease, with age, the level of LDL-cholesterol, smoking, the level of HDL-cholesterol, systolic blood pressure, family history of myocardial infarction, diagnosis of diabetes, and triglycerides all making significant contributions. As in other studies, low HDL-cholesterol emerged as a significant and independent risk factor for coronary disease, providing support for strategies aimed at correcting low HDL-cholesterol in addition to those aimed at reducing elevated levels of LDL-cholesterol.

  • Cardiovascular risk
  • Coronary heart disease
  • Cardiovascular risk engines

Introduction

Risk stratification is central to the management of vascular disease, and patients with a 10 year probability of myocardial infarction of ≥20% are deemed suitable for urgent intervention to reduce the impact of cardiovascular risk factors.1 Data from large epidemiological cohort studies have provided the raw data used to evaluate the contribution of individual cardiovascular risk factors to the overall level of cardiovascular risk. For example, the landmark Framingham study in the USA has provided a wealth of information within a database large enough to allow individual evaluation of the importance of hypertension,2 dyslipidaemias,35 and smoking,6 among many other factors. The data from the Framingham study have shaped the development of guidelines for the management of cardiovascular risk in Europe7 and in the USA.8 Many physicians worldwide use the risk calculation engine based on the Framingham cohort9 to identify patients in urgent need of intervention to reduce the likelihood of a morbid coronary event.

It is important, however, to recognize the limitations of individual observational cohort studies. The Framingham study, for example, is based on a relatively homogenous population of US citizens within a limited geographical area. As such, important variations between local populations may influence the prognostic value of some aspects of the risk calculation. It is important that additional cohort studies should be performed in different areas, to provide a means of risk stratification that is relevant to the populations under study. The PROspective CArdiovascular Münster (PROCAM) Study has analysed data from more than 26 000 subjects in Germany over a 25 year period. Data from PROCAM have thus been used to develop a risk calculator relevant to a northern European population, with adjustments for changes in risk factors for a number of precise geographical locations.

This review describes the background and design of the PROCAM study, and its associated risk calculation tool. Given the increasing recognition of the importance of low HDL-cholesterol in atherogenesis, special emphasis will be placed on the incidence and consequences of low HDL-cholesterol in the PROCAM cohort.

The PROCAM study

Background to the PROCAM study

The PROCAM population consists of >20 000 participants, aged 16–65, from 52 companies and government offices in the Münster and northern Ruhr areas of Germany.10 Recruitment of subjects without a history of myocardial infarction or stroke, and without ECG evidence of ischaemic heart disease, began in 1979 and closed in 1985. Follow-up was by questionnaire every 2 years, and morbid events and deaths were confirmed by examination of hospital records and/or eyewitness reports. The PROCAM cohort is still being followed up for heart disease, stroke, and mortality.

The PROCAM study has contributed importantly to progress in numerous fields of cardiovascular research. These include valuable clinical information on the importance of low HDL-cholesterol and elevated triglycerides, in addition to elevated LDL-cholesterol,11,12 novel risk factors such as Lp(a),13,14 other risk factors for myocardial infarction and stroke,15,16 the links between altered haemostasis and cardiovascular risk,17,18 the relationships between diabetes, cardiovascular risk factors, and cardiovascular death,19 and regional differences in cardiovascular risk profiles among participants in the Framingham and PROCAM cohorts (described in detail subsequently).20

The PROCAM risk algorithm

The PROCAM risk calculator is based on a scoring system for risk factors derived from 10 year follow-up data from the PROCAM study.10 The algorithm was initially developed and validated in 5389 men aged 35–65 at entry into the study, of whom 325 developed acute coronary events, and has since been extended to include female participants. A Cox proportional hazards model was used to construct the PROCAM algorithm. The risk factor variables used to construct the model, in decreasing order of importance, were age, LDL-cholesterol, smoking, HDL-cholesterol, systolic blood pressure, family history of myocardial infarction, diagnosis of diabetes, and triglycerides. The contribution of all of these risk factors to the overall model was statistically significant (P=0.018 to P<0.001).

A simple scoring scheme is used to replicate the model in practice to estimate overall cardiovascular risk. The continuous variables entered the model are lipid profiles (HDL-cholesterol, LDL-cholesterol, and triglycerides), systolic blood pressure, fasting blood glucose, and age. With the exception of age, these variables are subdivided into ranges with a pre-determined scoring system. For example, HDL-cholesterol ≤35 mg/dL contributes a score of 11 points, 36–37 mg/dL is given a score of 10 points, 38–39 mg/dL scores nine points, and so on, until an HDL-cholesterol level of 54–55 mg/dL scores one point, with higher levels incurring no score. The subject's current age minus 35 provides the score for this variable. Categorical variables are history of diabetes, history of myocardial infarction, receipt of antihypertensive therapy (all with a choice of ‘Yes’ or ‘No’), and smoking habit (‘Current’, ‘Former’, or ‘Never’). A score of zero is associated with ‘No’ or ‘Never’ with various positive scores associated with positive answers.

The sum of these scores provides the estimate of overall cardiovascular risk. The 10 year risk of an acute coronary event increased steeply at total PROCAM scores of above 50 (Figure 1). The usual definition of ‘high risk’ in cardiovascular terms (10 year risk of a coronary event of >20%) corresponded to a PROCAM score of 53. The projected risk of acute cardiovascular events derived from the PROCAM algorithm in this male cohort agreed well with the observed incidence of cardiovascular events in the study (Figure 2). One advantage of the PROCAM algorithm is that it identifies a discrete population of patients at high risk of an acute coronary event. Overall, three distinct populations, stratified on the basis of cardiovascular risk, were identified (Figure 3), among whom >20% of the cohort were at intermediate or high risk of a cardiovascular event within 10 years. Nevertheless, it is important to remember that only about one-third of all myocardial infarctions occur in the high-risk cohort. Although the incidence of myocardial infarction is lower in the moderate-risk and low-risk groups, the number of myocardial infarctions occurring within each are similar owing to the much larger number of patients within these groups (Figure 3).

Figure 1 Relationship between the overall PROCAM score and the risk of adverse cardiovascular outcomes.10 CHD, coronary heart disease.

Figure 2 Relationship between the estimated and actual coronary events* observed in the PROCAM study.10 Asterisk represents a projected 10 year incidence from the PROCAM algorithm of ≥40%. Global 10 year risk estimations for PROCAM scores of 0–20 and 21–28 were <1.0 and 1.0–2.0%, respectively.

Figure 3 Prevalence of acute coronary events within categories stratified for cardiovascular risk within the PROCAM cohort.10

Regional variations in cardiovascular risk

The design of cardiovascular risk engines should adequately reflect the pathophysiology of cardiovascular disease in the populations within which they are used. The relationships between cardiovascular risk profiles and prognosis in Germany were investigated using data from the PROCAM Study (n=8682) and the Augsburg (southern Germany) cohorts of the World Health Organization Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) Study from 1984 to 1985 (n=3001) and from 1989 to 1990 (n=2783).20

These three cohorts were essentially identical in terms of cardiovascular risk factors at baseline, with mean ages of 47–50 years, mean systolic blood pressures of 129–135 mmHg, incidences of diabetes of 2.6–5.1%, mean total cholesterol:HDL-cholesterol ratios of 5.2–5.3 in men and 3.9–4.1 in women, and an incidence of smoking of 32–35% in men and 18–25% in women. These cohorts were, therefore, considered to represent the local population with reasonable accuracy. The risk of coronary events increased steeply with age, as would be expected, with a cumulative incidence of fatal or non-fatal myocardial infarction in subjects aged 55–64 of 52.1 events/1000 subjects in men and 11.6/1000 subjects in women in the pooled MONICA cohorts when compared with 73.8/1000 subjects and 14.6/1000 subjects in the PROCAM cohort. Applying the Framingham risk formula to these populations consistently overestimated the actual risk, with the Framingham estimate markedly higher than the upper 95% CI for the observed value (Figure 4).

Figure 4 The Framingham risk algorithm markedly overestimates the risk of myocardial infarction in a population from Germany. Data from the MONICA Augsburg and PROCAM cohorts20 (45–54 and 55–64 represent age ranges).

Using the PROCAM risk calculator

A simplified version of the PROCAM risk calculator is available on the website of the International Task Force for the Prevention of Coronary Heart Disease.21 This group is affiliated to the International Atherosclerosis Society and steered by a group of leading physicians interested in the management of dyslipidaemia and cardiovascular risk. The data input screen is straight forward to use, with input for gender; age; levels of HDL-cholesterol, LDL-cholesterol, and triglycerides; systolic blood pressure; and whether there is a history of myocardial infarction, smoking, or diabetes, or a family history of myocardial infarction. An immediate estimate of the 10 year risk of myocardial infarction or coronary death is provided. Information on whether individual modifiable risk factors are within or outside normal limits is also given, along with suggestions for improving these where necessary. A further version of the PROCAM risk calculator runs on a CD-ROM.

Low HDL-cholesterol and cardiovascular risk within the PROCAM cohort

The importance of low HDL-cholesterol in determining cardiovascular outcomes was determined in 4559 PROCAM participants aged between 16 and 65.22 Within this cohort, 186 patients went on to develop coronary heart disease during 6 years of follow-up, including 165 confirmed myocardial infarctions. The severity of cardiovascular risk factors was greater, on average, in the patients without coronary heart disease when compared with those who went onto develop this condition, as indicated by lower HDL-cholesterol [1.2 mmol/L (45 mg/dL) vs. 1.0 mmol/L (40 mg/dL), P<0.001], higher LDL-cholesterol [3.8 mmol/L (147 mg/dL) vs. 4.6 mmol/L (176 mg/dL), P<0.001], higher triglycerides [1.5 mmol/L (135 mg/dL) vs. 1.8 mmol/L (163 mg/dL), P<0.001], and higher blood pressure (133/86 vs. 139/90 mmHg, P<0.01/P<0.05).

Between approximately 15 and 20% of men or women in this age group had low HDL-cholesterol, defined as <0.9 mmol/L (<35 mg/dL) or <1.2 mmol/L (<45 mg/dL), respectively. Subjects with low HDL-cholesterol had an approximately four-fold increase in the risk of coronary heart disease when compared with subjects with normal HDL-cholesterol and, conversely, the distribution of HDL-cholesterol levels was shifted towards lower values in patients with myocardial infarction (Figure 5). The incidence of coronary heart disease among individuals within the lowest tertile for HDL-cholesterol [<0.9 mmol/L (<35 mg/dL)] was in excess of 100 cases/1000 subjects when compared with 20–30/1000 subjects for the middle [0.9–1.4 mmol/L (35–55 mg/dL)] and upper tertiles [>1.4 mmol/L (>55 mg/dL)]. The incidence of coronary heart disease in patients stratified for both HDL-cholesterol and LDL-cholesterol is shown in Figure 6. Coronary risk was elevated in patients with low HDL-cholesterol, at any level of LDL-cholesterol. However, patients with combined low HDL-cholesterol and elevated LDL-cholesterol were clearly at higher cardiovascular risk when compared with patients with an abnormality of either lipid component alone. Consistent with these observations, levels of HDL-cholesterol exert a much stronger influence on cardiovascular risk in patients with high overall cardiovascular risk (90th percentile for PROCAM score) when compared with patients with the lowest cardiovascular risk (<90th percentile for PROCAM score) (Figure 7).

Figure 5 Levels of HDL-cholesterol in PROCAM participants with (MI+) and without (MI−) a prior history of myocardial infarction (unpublished results).

Figure 6 Relative contributions of LDL-cholesterol and HDL-cholesterol to coronary risk in the PROCAM study. To convert cholesterol values to millimole per litre, multiply by 0.02586. Reprinted from Assmann G, Schulte H, von Eckardstein A et al. with permission from Elsevier.22

Figure 7 Low HDL-cholesterol is most strongly associated with increased coronary risk in patients with high overall cardiovascular risk. Data from the PROCAM Study (unpublished results).

Discussion

The PROCAM study provides a well-validated means of assessing overall cardiovascular risk, based on commonly measured risk factors. The PROCAM risk calculator has some advantages over other risk engines. The PROCAM algorithm21 includes information on family history of coronary heart disease, triglycerides, and LDL-cholesterol within its calculations, whereas the Framingham calculator9 does not. In addition, the PROCAM risk calculator has been shown to more accurately reflect the prevailing coronary risk in a northern European population than the Framingham equivalent. Further, geographical localization of regional differences in cardiovascular risk profiles is possible and is underway to increase further the precision of the PROCAM risk calculator as a tool for cardiovascular risk stratification. A third risk calculator, based on the population of the UK Prospective Diabetes Study,23 is relevant to diabetic patients only and will not be discussed further here.

Low HDL-cholesterol was associated with an increased risk of coronary heart disease in the PROCAM study, consistent with the results of other analyses, including the Framingham study and the Helsinki Heart Study.2429 Each of these studies emphasized the independent nature of low HDL-cholesterol as an independent risk factor for adverse coronary outcomes, largely irrespective of levels of LDL-cholesterol, in multivariate analyses. In the case of PROCAM, both low HDL-cholesterol (P<0.001) and elevated triglycerides (P=0.018) were significantly associated with an increased risk of coronary events on multivariate analysis, taking into account other cardiovascular risk factors.10 Although combined hyperlipidaemias, where low HDL-cholesterol coincides with elevated triglycerides or LDL-cholesterol, confer the highest cardiovascular risk, the results of PROCAM and the other analyses described earlier support the concept of correcting low HDL-cholesterol in addition to other disturbances of lipid metabolism as a rational therapeutic strategy within an overall cardiovascular risk factor management programme.

Conclusions

Accurate risk stratification is essential for the effective management of cardiovascular risk. The PROCAM risk calculator provides a well-validated and straightforward method for risk stratification on the basis of a large cohort of European patients. The results of the PROCAM study support a role for low HDL-cholesterol as an important driver of cardiovascular risk, as has been found in other major observational and intervention studies.

Note added in proof

Please note that the studies cited in the present report were completed at different times, so that differences may be present in the absolute numbers of cases and events reported throughout the text.

References

View Abstract