Why Framingham? - Dr Anthony Wierzbicki
Multiple risk factors contribute to cardiovascular disease (CVD) risk but, apart from rare severe cases (e.g. severe young-onset hypertension), individual risk factors do not predict long-term risk very well.
Recourse is made to CVD risk calculators derived from epidemiological studies (e.g. Framingham)1 or primary care databases (e.g. QRISK)2,3 to give approximations of future risk.
All risk factor calculators agree that age, gender, smoking, total cholesterol: HDL ratio and BP are significant risk factors in all individuals.4 Age and gender are the dominant risk factors and others add comparatively less.
Diabetes, while part of original risk calculators, is now excluded as all patients with diabetes are considered high risk and the additional risk captured by ECG left ventricular hypertrophy is neglected for convenience.
Additional risk in some subgroups is added by further post hoc modifications e.g. ethnicity, family history of early CHD and obesity.1
Additional risk factors
The statistics behind the calculators and CVD risk models derived from these studies rely on the identification of independent risk factors.
In the original Framingham study, obesity and renal dysfunction were not significant as their effects were subsumed in either the lipid ratio or BP respectively.
Modern work has suggested that the lipid ratio and obesity are alternates for each other.5 Additional risk factors may not add significantly to prediction when added to the original scoring system.
Efforts should be made to minimise the number of risk factors used so as to reduce errors that cumulate rapidly with additional variables.
The Framingham calculator uses a basic risk marker set of high validity with minimal additions for subgroups (family history).
In contrast, QRISK includes deprivation scores, BMI, estimated GFR (eGFR), autoimmune disease and AF. All the co-efficients in these risk equations are subject to variation and the risk assessment also has to include the effects of assay variation and of biological variation. These tools only approximate risk.
Given this lack of reliability, calculators need to be validated prospectively in multiple populations. These data exist for both Framingham2, 6 and QRISK.2,7 However, in the QRISK validation studies only 25 per cent had complete risk factor profiles.2, 7,8
Framingham and QRISK differ in their endpoints. Framingham includes angina as well as the definite endpoints of CHD events or stroke, while QRISK relies on the accuracy of primary care cardiovascular risk registers for its endpoints, which includes transient ischaemic attacks (TIAs).2
Framingham is known to over-predict risk in modern European populations by about 30 per cent, while QRISK under-predicts by 0-12 per cent with significant discrepancies in the estimation of risk between genders, implying some bias in the underlying function.7,8
For Framingham this over-prediction can either be handled by re-basing the calculator9 using a more modern version, or in effect assuming that a 20 per cent risk actually includes the lower boundary of the confidence interval of the risk estimate.
It is worrying that 25 per cent of patients are re-classified by using QRISK as opposed to Framingham.7 No independent well-modelled confidence interval data exists for QRISK because the algorithm is not in the public domain.
Unlike Framingham, where the data are complete, QRISK relies on age and gender-based averages for missing data which are then interpolated.
Given that some data such as for HDL-C are >60 per cent interpolated - the QRISK calculator is effectively working using age and gender5 and partially correcting for HDL errors by simultaneous use of BMI - a variable known to substitute for lipid ratios.5
This could apply to many other interrelated variables (many associated with the metabolic syndrome) necessary to make QRISK comprehensive, all adding to 'accuracy' and sensitivity but losing precision due to reduced specificity.
Framingham is a well-established minimal model (and so the most precise). It relies on comprehensive accurate uniformly measured risk factor data, has complete and validated endpoints and has been validated in many cohorts.
As with all risk calculators the current Framingham model explains only about 70-75 per cent of the event risk and there may be a role for additional technologies that indirectly estimate the burden of atherosclerosis for patients at intermediate risk.
Risk calculators should be viewed as only screening tools. The best diagnostic investigations still need to be defined and the money found to pay for them to be used to ensure that only patients who actually require drugs receive them.
- Dr Wierzbicki is consultant in metabolic medicine and chemical pathology, Guy's & St Thomas' Hospitals, London
- Declaration of interests: Dr Wierzbicki has received grants and honoraria from various pharmaceutical companies. He was also chair of the medical committee of HEART-UK and served on the NICE technology appraisal group for ezetimibe and the guideline group for familial hypercholesterolaemia
Why QRISK? - Dr Mike Knapton
The evidence-base for managing CVD risk in a clinical setting was set out in the second Joint British Society Guidelines (JBS2) in December 2005. In 2008 the DoH published its proposals to develop a systematic risk assessment of patients aged 40-74 in its policy paper Putting Prevention First.
Both documents are predicated on the ability to adequately assess absolute risk in patients who have not yet presented with CVD and stratify that risk.
A key difficulty with Framingham is that it tends to overestimate risk in low prevalence populations. In particular, it fails to take into account ethnicity, although the modified Framingham algorithm does include family history and ethnicity.
The QRISK2 algorithm uses age, gender, smoking status, systolic BP and total cholesterol:HDL ratio. In addition, it considers BMI, family history of CVD in first-degree relative, Townsend deprivation score, current prescription of at least one antihypertensive agent, ethnicity and co-morbidities.
Published research indicates that QRISK will identify individuals more efficiently and equitably than Framingham.Validation of QRISK shows a better correspondence between the predicted and observed populations affected by CVD and therefore offers substantial advantages over Framingham.
However, to date there has been no head-to-head comparison.
It is clear from the development of QRISK that the recommendations of JBS2 have not yet been fully implemented and a considerable number of patients have still not had a comprehensive multiple risk factor assessment of risk of CVD.
Under these circumstances it seems a moot point which risk assessment algorithm you should use. Perhaps the priority should simply be to encourage a risk assessment to be undertaken at all.
There remains concern that both risk assessment algorithms underestimate risk, particularly in middle-aged women. By virtue of their age and gender middle-aged women tend to be assessed as low risk on the basis of a 10-year risk assessment.
Risk could be quantified over a longer period (say 30 years) or possibly lifetime risk. This would further enhance our ability to identify those who are most likely to benefit from treatment and avoid the risk of reassuring individuals inappropriately. This will be considered in some detail over the coming months as the Joint British Societies review their guidance.
Ultimately, individual risk management plans require effective communication of risk, behaviour change on the part of the patient and medical management.
QRISK has a number of benefits over Framingham. However, at the moment I think the priority is to ensure that all patients undergo a multiple risk factor assessment for CVD so that risk reduction can be targeted in the context of primary prevention.
- Dr Knapton is a GP in Cambridge
- Declaration of interests: Dr Knapton is associate medical director of the British Heart Foundation. The views expressed in this article are a personal view and do not represent the BHF position
1. Sheridan S, Pignone M, Mulrow C. Framingham-based tools to calculate the global risk of coronary heart disease: a systematic review of tools for clinicians. J Gen Intern Med 2003; 18(12): 1,039-52.
2. Hippisley-Cox J, Coupland C, Vinogradova Y et al. Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study. BMJ 2007; 335(7611): 136.
3. Hippisley-Cox J, Coupland C, Vinogradova Y et al. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2. BMJ 2008; 336(7659): 1,475-82.
4. Wierzbicki A S, Reynolds T M. Vascular risk screening: possible or too much, too soon? Int J Clin Pract 2009; 63(7): 989-96.
5. Gaziano T A, Young C R, Fitzmaurice G et al. Laboratory-based versus non-laboratory-based method for assessment of cardiovascular disease risk: the NHANES I Follow-up Study cohort. Lancet 2008; 371(9616): 923-31.
6. Liao Y, McGee D L, Cooper R S et al. How generalizable are coronary risk prediction models? Comparison of Framingham and two national cohorts. Am Heart J 1999; 137(5): 837-45.
7. Collins G S, Altman D G. An independent external validation and evaluation of QRISK cardiovascular risk prediction: a prospective open cohort study. BMJ 2009; 339: b2584.
8. Hippisley-Cox J, Coupland C, Vinogradova Y et al. Performance of the QRISK cardiovascular risk prediction algorithm in an independent UK sample of patients from general practice: a validation study. Heart 2008; 94(1): 34-9.
9. Brindle P, Emberson J, Lampe F et al. Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study. BMJ 2003; 327(7426): 1,267.
- Joint British Societies Guidelines on the Prevention of Cardiovascular Disease in Clinical Practice. December 2005.
- Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2 June 2008.