QRISK2 vs Framingham
Dr Peter Savill's editorial, ‘QRISK2 vs Framingham' (Practitioner 2008;252:5) provides a good summary of the QRISK2 CVD risk algorithm and rightly reminds readers about the approximate nature of CVD risk assessment. However, as a QRISK co-author, I cannot help but take issue with a few of his statements.
Regarding the Framingham risk score, Dr Savill states: ‘most GPs will be familiar with this algorithm and it is robust and well tested as long as appropriate adjustments are made (eg for family history and ethnicity).'
In my experience of teaching GPs on this subject, they may have heard of the name but most have little idea about all the different Framingham scores.
Most GPs are also probably unaware that the Framingham risk score recommended by NICE and JBS2 is not even a score published by the Framingham group, but rather two separate equations added together and adjusted by various modifying factors derived from completely separate population data. Consequently, this so-called Framingham score has no statistical credibility and can often come up with individual risks of >100%, which is difficult to explain to patients.
Different Framingham scores have been tested in different British populations and almost all studies have shown them to be poorly calibrated. Several studies have also found that the Framingham scores do not detect the additional CVD risk associated with social deprivation. Therefore I find the words ‘familiar', ‘robust', ‘well tested' and ‘appropriate' at odds with the evidence.
Dr Savill goes on to comment that QRISK2 ‘has only been validated in a single cohort' and requires ‘additional validation in other populations, particularly with different ethnic mixes.'
QRISK has been validated in 1.07 million patients from the THIN database (representative of 20% of British practices), and QRISK2 was tested in a separate sample of British practices (representative of another 60% of the population). This leaves only the remaining 20% of the population to test it in.
These validation studies are several orders of magnitude larger than others done before, and the test populations are entirely representative of the population that the score is designed to be used in: British general practice.
Dr Peter Brindle
GP, Bristol, R&D strategy lead, Bristol, North Somerset and South Gloucestershire PCTs and QRISK co-author
Dr Peter Savill replies:
I would like to thank Dr Brindle for his comments. My personal experience of GPs within my locality is that they are indeed familiar with the Framingham risk score, and in many cases it is integrated into a clinical system template to provide risk scores.
I accept that the finer details and statistical methods used to determine which Framingham equation is used may not be appreciated by those at the coal face in primary care. However, the statistical complexities of QRISK are likely to challenge most GPs; indeed, I believe the NICE guideline development group enlisted the help of statistical experts to unravel some of the intricacies of QRISK2.
I fully accept that QRISK2 appears to be a superior tool but gaining acceptance will take time, particularly as NICE currently recommends Framingham. However, I am sure that the THIN validation cohort will increase the acceptance of QRISK2 as the standard CVD risk calculator for the UK population.QRISK2 vs Framingham