Risk Estimation Works for Populations but not for Individuals
The QRISK-2 algorithm improves over the first version and represents a great leap forward over the Framingham equation because it is based on 16 million person years of cardiovascular events and has been validated against 2.22 million people, including a variety of ethnicities. This has allowed assessment of accuracy of the modelling parameters for prediction and improved accuracy when applied to a population. It also means that unlike other risk algorithms it is grounded in its population and does not suffer concordance discrepancies that reduce the validity of methods that attempt to mimic performance of other algorithms [2,3].
Unfortunately, when applied to individuals, all screening algorithms suffer the same problem. Intra-individual variation in lipid and blood pressure measurements mean that the tight estimates of the confidence intervals cited are over-optimistic and are mostly driven by regression dilution in the large sample employed  The estimates do not reflect the variation at the individual patient level. Many patients have only a single estimate of cholesterol. Modeling of the Framingham equation-based individual risks using data from published sources on variances shows that any estimated risk has a wide confidence interval meaning that when a patient is advised that their risk for example is 20%, the 95% confidence intervals for that estimate are ±6% (so 95% range is 14% - 26%) . Cardiovascular risk prediction also has many limitations . Consequently, whilst it is easy to identify a population at risk, it is not so easy to identify individuals at risk and risk estimation cannot be reduced to a production line process ignoring the role of detailed clinical assessment and significant medical experience.
Most cardiovascular events occur in people who have lipid and blood pressure results similar to the unaffected population. Thus targeting high -risk individuals actually has little effect on the overall burden of disease. It simply consumes resources. Instead, if the average cholesterol of the entire population was reduced by 0.5 – 1.0 mmol/L, which could be achieved by changes in diet, the use of plant stanols and/or the outcome- evidence based approach of low doses of statins (e.g. pravastatin 10mg/day), similar reductions in cardiovascular morbidity and mortality could be generated without the need for costly screening programs . Secondary prevention would then be applied in a similar way to anyone developing cardiovascular disease but this time using high dose potent statins as in the trials. This is the approach that actually underlies the new NICE guideline but the authors did not have the courage to state outright. If statins are the new aspirin for the 21st century then let us use them in the same manner.
Timothy M. Reynolds FRCPath: Professor of Chemical Pathology. Queen’s Hospital, Burton-on-Trent Staffordshire DE13 0RB
Adie Viljoen FRCPath: Consultant Chemical Pathologist. Lister Hospital, Stevenage Hertfordshire SG1 4AB
Patrick J Twomey FRCPath: Consultant Chemical Pathologist. The Ipswich Hospital Suffolk IP4 5BD
Anthony S.Wierzbicki FRCPath: Consultant Chemical Pathologist. St Thomas' Hospital, London SE1 7EH
1) Hippisley-Cox J, Coupland C, Vinogradova Y, Robson J, Minhas R, Sheikh A, Brindle P. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRisk2. Brit Med J 2008; 336 : 1475
2) Reynolds TM, Twomey P, Wierzbicki AS. Concordance evaluation of the coronary risk scores: Implications for screening. Curr Med Res Opin 2004; 20: 811-8
3) Wierzbicki A, Reynolds T, Gill K, Alg S, Crook M. A comparison of algoriothms for initiation of lipid lowering therapy in primary prevention of coronary heart disease. J Cardiovasc Risk 2000; 7: 63-73
4) Reynolds TM, Twomey P, Wierzbicki AS. Accuracy of cardiovascular risk estimation for primary prevention in patients without diabetes. J Cardiovasc Risk 2002; 9: 183-90
5) Greenland P, Lloyd-Jones D. Time to end the mixed and often incorrect messages about prevention and treatment of atherosclerotic cardiovascular disease. J Am Coll Cardiol 2007; 22: 2133-5
6) Reynolds TM, Mardani A, Twomey PJ, Wierzbicki AS. Targeted versus global approaches to the management of hypercholesterolaemia. J Roy Soc Health 2008; in press
Competing interests: None declared
Competing interests: No competing interests