Understanding the limits of coronary risk assessment tables
Dr Peter Brindle suggests ways to work around the weaknesses in
the Framingham heart study data and other risk estimation tools
There is increasing pressure for coronary risk scoring to be a part of primary care. The national service framework for coronary heart disease recommends that apparently healthy individuals at high risk of developing cardiovascular disease should be offered advice and preventive treatments, including blood pressure and lipid-lowering treatment, and aspirin.
The methods used to identify these people, such as the Joint British Societies guidelines published in the back of the BNF and the computer-based risk calculators, use age, sex, diabetes, smoking, blood pressure, total to HDL cholesterol ratio and left ventricular hypertrophy, to predict an individual's absolute risk of developing CHD.
These risk factors come from a regression equation derived from the Framingham study and published in 19911 (see box opposite).
Areas of differing CHD risk
There has been a 50 per cent drop in male CHD mortality since the 1970s, so the Framingham score may no longer
be applicable to modern practice in Britain.
There is evidence that Framingham overestimates the risk in Italy, Germany and Denmark, and a recent study has shown it overestimates CHD risk by about 50 per cent in a representative sample of British men3. This found that overestimation was highest in areas of Britain, such as the south, where CHD risk was lowest.
Black and minority ethnic groups
CHD rates vary between ethnic groups and guidelines suggest multiplying the Framingham score by 1.5 to compensate for the presumed higher risk of south Asian groups.
Unfortunately, this figure is based on mortality by country of birth so does not reflect the risk in second-generation immigrants. It also fails to recognise the variation in risk factor levels between south Asian groups.
There are no official recommendations for people of African or Caribbean origin, who, compared with south Asian and White populations, have a relatively low rate of CHD but a high risk of stroke. A recent study has recommended using lower thresholds in these groups for the management of hypertension4.
Family history of CHD
As CHD is so common it is difficult to define what is meant by a 'family history'. One working definition is having an affected male first-degree relative under the age of 55 or female relative affected under 65, and compensating for this by multiplying the individual's Framingham risk by 1.5.
Clearly, different degrees of family history are possible, and the message is to recognise that your patient is of greater than average risk and to lower your threshold for intervention.
All or nothing?
Smoking and diabetes are included in the score as binary variables, meaning the five-a-day smoker is treated the same as a 40-a-day smoker, and the well-controlled diabetic the same as the chaotic. Clinical common-sense is required.
Framingham has been criticised for emphasising age too much. A 74-year-old man with relatively low risk factors ends up on blood pressure or lipid-lowering medication because of his age while the 32-year-old man with a cholesterol level of eight is denied treatment because his 10-year risk is low. This is disconcerting for clinicians used
to treating conditions such as hypertension and hyperlipidaemia at specific, but often arbitrary, thresholds of the individual risk factor irrespective of absolute risk.
The obese and sedentary
Although these are absent from the Framingham score, we know that, all other things being equal, obese and inactive people are at greater risk of CHD than the slim and active. Once again, we have to rely on our clinical judgment to adjust for these factors rather than a chart or computer.
Implications for prescribing
If all that matters is reducing cardiovascular risk there is already much evidence suggesting we should offer more preventive treatment to people of lower risk rather than restricting it to those at very high risk.
However, in reality we will prescribe what workload and cost permits and what our patients are prepared to take. Current guidelines recommend offering blood pressure treatment to those at =15 per cent 10-year risk and statins at =30 per cent, and these thresholds have been determined by affordability.
It is rarely acknowledged that the large majority of coronary events occur in those with a predicted risk that is below the 30 per cent 10-year risk threshold3. These thresholds are likely to be lowered in due course as statins come off patent and become cheaper.
So the concern that Framingham may considerably overestimate CHD risk in modern Europe may inadvertently make up for its possible underestimation in high-risk groups and the rather conservative treatment thresholds we have for the rest of us.
The SCORE guidance
A recent system derived from European datasets to address the problem of using a single North American score throughout Europe is being promoted as an alternative to the Framingham score2,5.
Although not ready for routine clinical use until early 2004, SCORE will consist of a risk score that can be adjusted to the incidence of cardiovascular disease in different countries.
However, within Britain, Framingham overpredicted CHD risk by about 71 per cent in the southern sample of the British regional heart study and by 28 per cent compared with the observed risk in Scotland.
So far, it is unclear how SCORE will take account of such regional variations within countries. Furthermore, SCORE fails to include ethnicity, and like Framingham it does not take account of other risk factors not included in the risk score such as family history, obesity and sedentary lifestyle. It also uses the same binary smoking variable.
However, SCORE differs from the Framingham score in that it does not attempt to predict risk in diabetics and also in its choice of cardiovascular disease death as the endpoint instead of coronary heart disease events that includes angina.
Although clinicians may be more interested in cardiovascular disease, which includes stroke, this focus on mortality reduces its potential usefulness.
Doctors are as interested in the risks of disabling morbidity as they are in outright death.
In the meantime
Whether SCORE will overtake Framingham to become the dominant risk-scoring tool for the primary prevention of cardiovascular disease remains to be seen.
They both share most of the same difficulties of using historical data from specific populations to predict the future risk of individuals.
However, as long as we are aware of the limitations of such techniques, risk scoring is extremely useful for prioritising treatment but the precise risk estimate obtained can give an illusion of certainty about the future.
Fortunately, the gaps in the risk-scoring approach can be at least partially filled by the confident use of our clinical judgment.
The Framingham Heart Study
The Framingham study was started in 1948 and is a large prospective study based in what was the predominantly white and middle-class town of Framingham in Massachusetts, USA. Risk factor information from 5,573 men and women was collected between 1968 and 1975.
Framingham-based risk scoring methods have become dominant because of the consistent methods used in the study and the relatively large number of female participants an unusual feature of cardiovascular cohorts at the time.
Alternative risk-scoring tools such as the Dundee risk-disk, the British Regional Heart Study score and the Pocock score never really achieved popularity. However, the recently announced SCORE algorithm, derived from a number of European datasets, is being promoted as a serious alternative to the Framingham score2.
1 Anderson KM et al. An updated coronary risk profile.
A statement for health professionals. Circulation 1991;83:356-62
2 Conroy RM et al. Estimation of 10-year risk of fatal cardiovascular disease in Europe: the SCORE project.
Eur Heart J 2003;24:987-1003
3 Brindle P et al. The Framingham score overestimates coronary risk in British men. A prospective study. BMJ 2003 (in press)
4 Cappuccio FP et al. Application of Framingham risk estimates to ethnic minorities in UK and implications for primary prevention of heart disease in general practice: cross-sectional population-based study. BMJ 2002;325:1271
5 De Backer G et al. European guidelines on cardiovascular disease prevention in clinical practice. Eur Heart J 2003;24:1601-10