This site is intended for health professionals only

At the heart of general practice since 1960

How useful are

Framingham

charts

in general

practice?

The new emphasis on CVD, rather than CHD, risk has put the Framingham charts in the spotlight ­ Professor Hugh Tunstall-Pedoe looks at the pros and cons of risk scoring, while Dr Stephen Willott gives a GP's perspective

Cardiovascular risk scoring has one basic function: to advise who should be a priority to receive medical time and medication to prevent a cardiovascular event, in an environment where resources for prevention are limited. People at the highest risk stand to benefit most and are the most efficient group to target.

Of course the top priority for prevention is people who already have vascular disease. The disease-free and symptom-free come second in that respect, as their risk is not so high.

Good clinical histories and good records identify the highest-risk patients without any scoring. Now thresholds for treatment are coming down, GPs will increasingly have to consider treatment in the apparently healthy and this will require more risk scoring.

Risk scoring vs clinical judgment

Some GPs with years of experience of risk scores can estimate the results of the calculations, but for other GPs the charts can perform a valuable function by contradicting years of teaching of clinical myths.

The commonest myth is that high elevation of one risk factor matters more than modest elevation of two or three together. The latter can be much more serious, is much more common, and accounts for a much larger proportion of future cardiovascular catastrophes.

In the past, GPs thought that mild elevation was within the normal range so they tended to ignore it and go for the rare birds like systolic blood pressure over 200mmHg and cholesterol over 10mmol/L, missing the fact that high risk affects a significant proportion of people who appear very ordinary until you calculate their risk scores.

Another advantage of risk scoring is that your practice can get a bigger picture of what is going on, and prove that resources are being distributed fairly and transparently according to need and not according to

arbitrary judgments.

After all the high-risk strategy is discriminatory, and it is preferable to be able to show that the discrimination is based on accepted guidelines and not just on who you like.

Any risk score can only grade people according to the factors that are in the formula, which tend to be the three or four classic risk factors.

Of course, a risk score developed in one population in one decade will not be correctly calibrated when used in another country some decades later.

However, all the common cardiovascular risk scores will agree with each other very well as to who is high risk, who is low risk and who is somewhere in-between. And it is the prioritisation that matters.

In that respect arguments about accuracy are rather like disputes between mediaeval theologians ­ they become metaphysical, and seem to me to miss the point.

I developed some charts in the 1990s ­ the Dundee coronary risk-disk ­ which ranked people from 1-100 in their likely position in the bus queue for having a coronary. The concept is very valid and the ranking by most other competing scores would have been in good agreement.

Absolute risk and risk predictions

Before statins arrived, risk scores such as

the Dundee risk-rank were used to motivate behavioural change rather than to initiate medication. Subjects' cardiovascular risk was compared with their own age and sex peer-group by relative risk, to motivate this change, so age and sex were taken out of the score. But medications have definite costs and definite risks associated with them so age and sex were put back into the formula and risk was expressed in real terms, such as mortality or incidence rates, to establish thresholds across wide population groups below which medication would not be justified for these reasons.

The word 'absolute' means real risk and not relative risk, as 'absolute' and 'relative' are opposing terms in philosophy, but

there is nothing 'absolute' about a risk score result, as it is a crude assessment or estimate of risk. It would be amazing if the people classified in Britain as being at 15 per cent 10-year risk in the Framingham score actually did have that risk.

Most studies (except a recent one dating back to the 1970s in the west of Scotland) suggest the Framingham score overestimates absolute risk in modern populations.

However, that does not really matter as the decision on treatment thresholds is quite arbitrary ­ we have recently heard that the threshold is going to come down substantially in England and Wales. What does matter is not what the threshold is, but that those above the threshold have a good chance of being assessed and treated in preference to those below it ­ anything else would be unfair.

My reasons for deploring the horrible term 'risk prediction' are the same. It

implies that doctors have the power of prophecy like gypsies in a fairground with a crystal ball. They are assessing risk in the here and now and making an estimation to help with prioritising treatment. The accuracy of these assessments in real terms will never be known unless they were to have thousands of patients followed for decades without any preventive intervention!

Both 'absolute' and 'risk prediction' imply supernatural powers which are currently denied to us.

Be careful over 10-year risks. A 10-year risk of 10 per cent does not mean 1 per cent in the first year, as risk increases by compound interest of about 15 per cent a year, doubling in approximately five years. Risk in the first year will be much less than you might expect.

Framingham in practice

I do think the Framingham score underestimates the effect of age so that it exaggerates it in younger people and minimises it in older people.

But this may not be a bad thing ­ actually it seems to be what clinicians want because otherwise age becomes the overwhelmingly dominant risk factor and the others, apart from sex-group, much less so.

I also think risk charts rather than computerised risk scores may be doctor- and nurse-friendly but they are not patient-friendly ­ if you are near a boundary line your risk can suddenly double or halve through random variation in repeated readings and that must be very disheartening.

Statistically speaking it is not possible to modify Framingham without reworking it, although results can be 'tweaked'. Adding new risk factors has a very disappointing effect in improving prediction in risk scores once you have incorporated three or four.

Deprivation and ethnicity

Framingham does not include social deprivation and ethnicity. And so the Framingham score is 'unfair' when comparing people with different levels of deprivation or ethnicity, because it is blind to them. I am currently involved in discussions on how to address this problem in Scotland.

The deprivation question does not seem to be confined to the 'submerged 10th of the population', as there is a powerful social gradient right across with a big jump in risk even between the most affluent fifth of the population and the next group along.

These problems are not a reason for rejecting risk scoring or the Framingham score, but we cannot ignore them in planning for the future.

Scoring should be made fairer rather than abandoned altogether, which would result in greater unfairness.

If you are adopting a high-risk approach, you have to define high risk and because high risk is multifactorial you need a score.

The alternative is the mass-population

approach ­ put the drugs in the drinking

water. The nearest approach we have had

to that was the polypill proposal but even there high risk was incorporated to the extent that the polypill was for everyone over the age of 50.

It was a challenging proposal, but was not liked by doctors who prefer to target medication, and that means risk scoring.

Framingham under pressure

· June 2004 Two studies show Framingham substantially underestimates risk of CHD in diabetes

· September 2005 SIGN-commissioned study shows Framingham underestimates relative risk of deprived patients by 300 per cent

· October 2005 Study presented to NICE suggests adjusting Framingham could increase patients eligible for statins seven-fold

· December 2005 JBS recommends GPs assess all patients over

40 for CVD risk

· February 2006 UK experts develop a new online risk calculator

to allow GPs to adjust Framingham for different ethnic groups

Hugh Tunstall-Pedoe was recently made emeritus professor of cardiovascular epidemiology at the University of Dundee ­ he developed the Dundee coronary risk-disk

Competing interests None declared

A GP's view

GP Dr Stephen Willott describes how some GPs on the ground are trying to adapt risk scoring to their patient list

When you start to hear that the recommended risk assessment tools may be contributing to health inequalities, it makes you wonder.

Indeed the conclusion of Dr Peter Brindle's powerful paper looking at how Framingham faired for different groups was that deprived people were less likely to get treatment to prevent heart disease.

I know some GPs use other risk tools such as www.riskscore.org.uk but again this is based on data outside of the UK. Some GPs already simply add another 10 per cent to the 10-year risk estimate for south Asians. While this may be the right approach, clearly it needs to be more scientific/evidence based.

The new JBS 2 guidelines are to be welcomed but they seem to have missed the opportunity to highlight the necessary adjustments for both deprivation and ethnicity.

All that is said (in the appendix!) is the risk charts 'seem to underestimate CVD risk in some ethnic minorities...in people originating from the Indian subcontinent it is safest to assume that the CVD risk is higher than predicted from the charts

[1.4 times]'.

How we adapt the risks to take into account family history also needs to be clarified, but is likely to stay as a 'multiply standard Framingham risk by X amount' where X at the moment is 1.5. But I suspect this underestimates things and it should perhaps be X2 ­ but then it all depends on what you call a family history.

Thankfully, NICE, SIGN, the National Screening Committee and Connecting for Health are already all considering the problem.

I, like many others, hope it will not be long before the QOF addresses primary prevention but until then GPs have little means of altering primary care activity.

Perhaps we could all look at those in the deprived areas on our lists who are at increased vascular risk (but who currently score below the threshold for action) with renewed priority, to help reduce the inequalities rather than further increase them.

I'm sure we should be lowering the threshold for treatment for those with risk factors from a deprived area but as yet we have no idea how to go about this. Patients with diabetes at an arbitrary age of 40 or so are now generally included as secondary prevention so don't need risk scoring.

In the meantime, each clinician must rely on clinical judgment and look wider at 'whole person risk' including ethnicity and recognise the influence of the non-Framingham risk factors.

Stephen Willott is a GP specialist in public health and CHD clinical adviser, Nottingham City PCT

Competing interests None declared

Rate this article 

Click to rate

  • 1 star out of 5
  • 2 stars out of 5
  • 3 stars out of 5
  • 4 stars out of 5
  • 5 stars out of 5

0 out of 5 stars

Have your say