What is the background to this?
NICE guidance states that statin therapy is recommended for the primary prevention of CV disease for adults who have a 20% or greater 10-year risk of developing CVD. Predicted 10-year risk also influences decisions about other treatments, such as management of hypertension. National guidelines such as those produced by NICE and SIGN, and guidelines produced by learned bodies such as JBS, all advocate estimating the cardiovascular (CV) risk of people who do not have established disease using tools which consider several risk factors. This approach can be traced back more than 30 years.
But as Niels Bohr said: “Prediction is very difficult, especially about the future”. What if the prediction tool we use either overpredicts or underpredicts risk? We could end up either over or under treating lots of people. Many prediction tools use data from the Framingham study – a population study in a mainly white, affluent population in New England measured at a time when CVD was at its peak in the USA. Data from this study are the basis of, for example, the JBS tables printed in the BNF and the program used by EMIS. However, the population of patients on which these tools were derived is are perhaps different from the UK population. The 2007 SIGN guidance recommends the ASSIGN risk assessment tool, which is based on a cohort of Scottish patients and includes deprivation as a risk factor.
QRISK is a new CVD risk scoring system that was specifically developed for the UK. It is based on a large UK general practice database (QRESEARCH). A validation study of QRISK was first published in 2007 It seemed to show that QRISK is better than either ASSIGN or Framingham at predicting cardiovascular risk in a random sample (a third) of the QRESEARCH cohort (the remaining two-thirds of the cohort were used to derive the QRISK score). There was considerable discussion about this and the value of QRISK. The QRISK authors were concerned that because the validation sample was drawn from the same cohort as that use to derive the tool, there might have been what they call a “home advantage”. They therefore decided to carry out a second validation study, based on an independent UK general practice database, THIN.
An editorial accompanying this study explains that the two main measures by which a risk prediction tool should be judged are calibration and discrimination. Calibration relates to how close the predicted risk is to the observed risk. More importantly, discrimination is the ability of the tool to differentiate between people who will have an event and those who will not, over a defined period of time (often five to ten years).
What does this study claim?
The QRISK authors generated QRISK and Framingham scores for each individual patient (n=1,072,800) in the THIN cohort and the QRESEARCH cohorts and also looked at the actual CV event rates in each cohort.The authors state that QRISK performed better than Framingham for every discrimination and calibration statistic in both cohorts. Framingham overpredicted risk by 23% in the THIN cohort, while QRISK under-predicted risk by 12%. The ratio of predicted:observed events in the THIN cohort using QRISK was 0.90 for women and 0.87 for men (0.88 overall), and using Framingham it was 1.10 for women and 1.32 for men (1.23 overall). They conclude that QRISK is better calibrated to the UK population than Framingham and has better discrimination. The say that the results suggest that QRISK is likely to provide more appropriate risk estimates than Framingham to help identify patients at high risk of CVD in the UK.
It is clear that the QRISK score was better calibrated to the UK population included in the THIN database than Framingham. But, as the accompanying editorial points out, the Framingham investigators have previously described a method of calibrating the Framingham score for different populations. The editorial states that the differences between summary discrimination measures reported for Framingham and QRISK in this study are relatively small. The accuracy of discrimination using QRISK, at the 20% 10–year CVD risk treatment threshold, is likely to be quite similar to that from Framingham and ASSIGN.
The NICE guidance on lipid modification has been delayed, since the equation used to estimate cardiovascular risk is central to the guideline and the recommendation in the draft guideline is to use the Framingham equations. In the light of the validation study discussed here, NICE has asked its guideline development group to reconsider their recommendations on risk estimation and to seek advice on technical issues from independent experts.
Health professionals should be aware of the ongoing debate, and also that Framingham-based tools may over-predict CV risk in some sections of the UK population, but not others, such as those in high risk groups (e.g. socio-economically deprived, people of South Asian descent, those with a family history of CV events, etc). Even with these caveats, as the draft NICE full guideline on lipid modification says – estimates of CVD risk derived from equations are not an exact science but are better than clinical judgment alone for the estimation of CVD risk.
Of course, health professionals need to take into account patient circumstances and wishes. It would be foolish to have an iron rule that (whatever tool is used) someone with a 19.9% predicted risk can never receive prophylaxis, but someone with a 20.1% risk must always receive prophylaxis. The most important thing is to correctly use a validated tool – be it Framingham, ASSIGN or QRISK as a basis for discussion with patients and not to treat on the basis of individual risk factors (eg an isolated high blood pressure or isolated high cholesterol). As the editorialist to the QRISK paper writes: “risk prediction, like your tax return, is difficult to do in your head”