Cookie policy notice

By continuing to use this site you agree to our cookies policy below:
Since 26 May 2011, the law now states that cookies on websites can ony be used with your specific consent. Cookies allow us to ensure that you enjoy the best browsing experience.

This site is intended for health professionals only

At the heart of general practice since 1960

Is it time to turn our back on Framingham?

Framingham equations overestimate risk and should be replaced by QRISK, argues Professor Julia Hippisley-Cox. Professor Paul Durrington counters that there are good reasons for continuing to use it in a modified form for now, counters

Framingham equations overestimate risk and should be replaced by QRISK, argues Professor Julia Hippisley-Cox. Professor Paul Durrington counters that there are good reasons for continuing to use it in a modified form for now, counters

Yes

Estimates of cardiovascular risk work best in the type of population from which they were derived. The Framingham equations were derived from North American populations from the 1960s to the 1980s when CHD was at its peak.

This means they overestimate risk today by about 100% in southern Europe and by 50% or more in northern Europe, including the UK.

Overall, Framingham is likely to overestimate risk in affluent white males and underestimate risk in those from deprived areas, south Asian groups and women. To my knowledge, the 1991 Framingham equations are only used in England, Wales and New Zealand. They have been dropped by SIGN in Scotland and by the European guidelines, and have never been seriously considered in the US.

QRISK is a new CVD risk score that, like Framingham, estimates the risk of developing CVD over the next 10 years.

But unlike Framingham, QRISK – now on its second iteration QRISK2 – has been specifically developed by UK GPs for GPs using UK QResearch data from EMIS.

QRISK contains many of the traditional risk factors included in Framingham, such as age, sex, total cholesterol-HDL ratio, blood pressure, diabetes and smoking status. But it also contains additional risk factors, including ethnicity, deprivation, BMI, CKD, atrial fibrillation and family history of premature CHD in a first-degree relative under 60.

Framingham equations do not include socioeconomic status and underestimate risk in people who are relatively socially deprived and south Asian populations.

QRISK2 has been modelled on a vast quantity of real-world general practice data and although most patients have BP and BMI recorded, not everyone has a total cholesterol-HDL measurement. We used a statistical technique called multiple imputation that lets us model risk more precisely in the presence of missing data.

The validation of QRISK1 and QRISK2 was published in Heart earlier this year1.

An accompanying editorial says the main measures by which a risk prediction tool should be judged are calibration – how close the predicted risk is to the observed risk – and discrimination – its ability to differentiate between people who will have an event and those who will not. QRISK2 outperformed Framingham on both.

I've heard claims that using Framingham will prevent a greater number of CVD events, but this is not correct – especially if you are female, from certain ethnic groups or a deprived background. QRISK2 is more efficient at identifying higher risk populations and therefore the number of events prevented by intervention.

There have also been criticisms that the increased accuracy of QRISK2 compared with Framingham is because it includes patients aged 70-74 – but QRISK2 is actually more accurate than Framingham at all ages.

The current NICE recommendation is to take the risk of CHD using Framingham equations and add it to the risk of stroke, which is double-counting since CHD risk is not independent of stroke risk.

The result is then multiplied by 1.4 if the patient is a south Asian male and 1.5 if the person has a family history of premature CHD. If they have both, then both multipliers are used.

This modification was not tested or validated and, sadly, it gives risks of more than 100%, which is logically impossible.

QRISK2 has better discrimination and calibration than the NICE modification of Framingham, which will underestimate risk in women and people from deprived areas. QRISK2 will be updated periodically to reflect population changes and the latest evidence – yet another advantage over the static data used in Framingham.

Professor Julia Hippisley-Cox is a QRISK researcher and professor of primary care at the University of Nottingham. She would like to thank Dr Peter Brindle, a GP in Bristol and R&D lead for Bristol, North Somerset and South Gloucestershire PCTs, for his help with this article

No

Few would deny Framingham has its faults and QRISK2 has some theoretical advantages. But NICE has recommended that for the present we continue to use Framingham, preferably somewhat modified from its original version. And there are very good reasons for that.

NICE did originally propose to use QRISK, but external referees expressed major concerns that its adoption would be premature. In my opinion there are three important reasons for that decision.

First, the claim that Framingham overestimates risk should be qualified. It is likely to be true in older men, which is why Joint British Societies (JBS) guidelines recommend risk calculation only up to 70 years.

It is not clear why Professor Hippisley-Cox and colleagues used a population up to 75 years, but a lot of the apparently more accurate risk prediction is as a consequence of including these older people.

Second, the real issue is how best to prioritise people for statin and blood pressure treatment at the age when GPs might reasonably begin screening for CVD risk. In this respect, the performance of QRISK2 is very worrying.

For example, in the latest report from Professor Hippisley-Cox's group2 in the age group 50-54, QRISK2 identifies only 4.75% of men and 1.05% of women as at 20% or greater 10-year CVD risk, whereas Framingham identifies 19.7% of men and 4.15% of women.

We know that the fewer people who receive statin or antihypertensive therapy, the smaller the reduction in incidence of CVD in the population.

This is because the largest number of people who have CVD events – and thus in whom it can be prevented – are those whose risk is identified as 20% or more by Framingham.

In our study, using Framingham was almost as effective as treating the whole population. It does not matter whether those selected for treatment are at exactly 20% 10-year risk or more as long as they are close to this threshold.

So Framingham looks like a better bet in terms of the number of CVD events prevented. There is no point in all this effort by GPs and their patients if it is to have only a negligible effect on CVD incidence. In terms of priority, it would seem sensible to start screening in middle age.

Third, QRISK2 is based on incomplete data from GP registers. In particular, lipid data were available in only about a quarter of patients and CVD events were collected over only six years or so.

The accuracy of modelling for missing data over 10 years requires careful validation. Under-estimation of CVD incidence might make Framingham more accurate than QRISK.

Furthermore, modelling for measured variables like lipids and blood pressure will have the effect of attributing more risk to categorical variables such as age, ethnic background, gender and socioeconomic group.

These make an immutable contribution to risk. So the effects of treating cholesterol and high blood pressure are underestimated.

NICE has recommended that a modified Framingham be used. Framingham allows CVD risk to be calculated from age, serum cholesterol, HDL cholesterol, blood pressure, gender and smoking history.

NICE recommends increasing the calculated risk if there is an adverse family history or the patient is of Indo-Asian descent.

The JBS charts are already published in the BNF and their Cardiovascular Risk Assessor computer programme will be available on the websites of HEART UK, the British Hypertension Society and the British Heart Foundation.

The BNF clearly states that risk assessed from the charts should be multiplied by 1.5 if the patient has a family history of CVD or is of Indo-Asian descent.

Clearly we want to develop methods suited to the British population, to be used in the context of evidence-based guidelines agreed by GPs and hospital specialists. In the meantime, the modified Framingham methodology is the devil we know and should continue to be used in practice.

Professor Paul Durrington is professor of medicine at the University of Manchester and Heart UK representative for the JBS guidelines

Statin tablets

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