QRISK – a new CVD risk score
Most clinicians currently calculate a patient's risk of cardiovascular disease (CVD) with reference to a set of risk charts or a computer programme. These invariably use the Framingham dataset to derive the absolute risk of a cardiovascular event, usually over a 10-year period. It is accepted that there are limitations when applying the Framingham dataset to modern Western populations; notably the lack of ethnic diversity and the high incidence of CVD in the index population.
Risk charts, such as the widely used Joint British Societies' charts,1 do not include social deprivation, BMI and current treatment regimens and make only crude adjustments for ethnicity and family history. The recently developed Scottish risk calculator ASSIGN2,3 goes some way towards correcting these shortcomings by including social deprivation.
There is now a new risk score: QRISK.4 It has been developed from a UK general practice population open cohort study. Patients were identified from the QRESEARCH database, which extracts routinely collected data from the EMIS clinical system and includes census ethnicity and social deprivation data.
Participating GP practices were randomly assigned to either a derivation or validation cohort.
The derivation cohort consisted of 1.28 million patients aged 35-74, registered at 318 GP practices between 1995 and 2007, who did not have CVD or diabetes. The validation cohort consisted of 610,000 patients from 160 practices.
The study recorded any diagnosis of CVD and collected risk factor data such as age, sex, smoking status, systolic blood pressure, total cholesterol: HDL ratio, presence of left ventricular hypertrophy, BMI, family history of CHD in a first degree relative <60 years of age, Townsend social deprivation score, area ethnicity and existing antihypertensive treatment.
QRISK, a CVD risk algorithm, was developed in the derivation cohort. It was then applied to the validation cohort to test accuracy and performance against traditional Framingham calculations and the newer ASSIGN score. The final algorithm omitted left ventricular hypertrophy and area ethnicity as neither added any significant weighting to the final score and left ventricular hypertrophy was rarely recorded.
In the validation cohort the observed 10-year risk of a cardiovascular event was 6.6% (95% CI=6.48-6.72) and 9.28% (95% CI=9.14%-9.43%) in women and men respectively. The risk scores tended to overestimate risk to some degree, but this was less marked with QRISK (0.4%) than Framingham (35%) or ASSIGN (36%).
Fewer patients aged 35-74 were classified as high risk (? 20% risk of CVD over 10 years) with QRISK (8.5%) compared with Framingham (13%) and ASSIGN (14%). When applied to the 2005 UK population, this equates to 3.2 million patients with QRISK, 4.7 million with Framingham and 5.1 million with ASSIGN.
Therefore the novel risk score QRISK performs better for the UK population than the Framingham or ASSIGN models and could be a more accurate tool for GPs to calculate CVD risk and populate risk registers, which are likely to come to the fore in the near future.
The algorithm could be embedded in clinical systems to provide a rapid and contemporaneous risk score for patients. It is particularly encouraging to see a reduction in the number of patients calculated as being at high risk of CVD using the QRISK model, as the economic burden of managing such patients is considerable.
However, there are wider issues around risk calculators, such as treating otherwise healthy people based on arbitrary thresholds, which could divert resources aimed at treating established disease and funding research to develop treatment options.
The assessment of risk is often a clinical judgment.
A sedentary, obese smoker will be at high risk of CVD and should improve their lifestyle rather than pondering over a CVD risk calculator and leaving with a clutch of antihypertensives, a statin and aspirin. So yes, risk calculators have a place, and QRISK is certainly an improvement on Framingham, but the whole concept of risk, risk registers and treatment thresholds must be viewed as an adjuvant to clinical judgment and not as an end in itself.Author
Dr Peter Savill
BSc MB BS PGDipCard
GPSI Cardiology, Southampton