Cardiovascular risk tables

Abstract

Estimating risk is not the problem, using it to tailor treatment to individuals is

In the linked study, Hippisley-Cox and colleagues develop and validate the second version of the QRISK cardiovascular disease risk algorithm (QRISK2), an attempt to more accurately estimate cardiovascular risk in patients from different ethnic groups in England and Wales.¹

The advent of the first Framingham risk tables in the early 1990s was a challenge for most doctors. Since the second world war the management of cardiovascular risk has been part of the core business of general practice, but the single risk model dominated. Hypertension, diabetes, and hypercholesterolaemia were islands, each with its own experts fighting for bigger kingdoms by pushing for ever stricter boundaries and demanding more attention.

Framingham taught us to look at the different risk factors, and provided a major lesson: a cumulative average risk could be more important than one peak. Yet soon the extrapolation of these US tables to European populations seemed to overshoot the real risk in these groups.^{2 3} The SCORE tables used the same risk factors to calculate corrected European cardiovascular mortality.⁴

More recently the ASSIGN⁵ and now the QRISK tables tried to incorporate some other known risk factors, especially deprivation and family history. Again, a major step: for several decades the medical community has had to face the troubling fact that cardiovascular morbidity and mortality are strongly and independently related to deprivation.^{6 7} If we ignore this we overestimate the risk for rich people (and overtreat them) and underestimate that for poor people. It’s probably naive to think that we can close the gap in cardiovascular risk just by giving more statins to poor people.

If epidemiologists could estimate cardiovascular risk accurately, would it solve our problems in managing our patients? Not at all. Risk calculation itself is based on evidence. However, using risk calculation in managing patientsrelies on consensus. When does a “risk” become a “high risk”? At what moment does a high risk justify starting lifelong drug treatment? The SCORE tables are useful, but when the European Guidelines tried to implement these tables and defined a 5% risk of death within the coming 10 years as high risk (comparable to a 20% risk in the Framingham tables),⁸ it led to an enormous medicalisation of many healthy elderly people, as proved by the Nordic Risk Group.⁹ Nearly all Norwegian men aged 60 years and older and allwomen aged 65 years and older were classified as at “high risk”—in a population with one of the highest life expectancies in the world.

To use an absolute risk score as a threshold for starting drugs is dangerous and not evidence based. It is therefore surprising that the recent NICE guidelines strongly recommend statins for anyone with a cardiovascular risk score of 20 or more in the Framingham tables.¹⁰

Age is such an important risk factor for developing cardiovascular problems within the next 10 years that all risk tables are misleading. Becoming older is by far the strongest predictor for morbidity and mortality—this is a biological fact. By looking at the risk tables, anyone can see what happens: by age 65, a large group has reached the 20% risk threshold, and lipid lowering drugs are prescribed for the rest of their lives.

A non-smoking man of 70 with a systolic blood pressure of 130 mm Hg and a total cholesterol concentration of 5 mmol (far below the median cholesterol concentration in most European countries) is at high risk according to the SCORE criteria. Unfortunately most of the trials of statins include only a few people older than 70.¹¹ The PROSPER trial, which specifically looked at this elderly population, showed that the primary composite endpoint (cardiovascular death or non-fatal infarction orcerebrovascular accident) was lowered by only 15% (48 people have to be given statins for three years to prevent one event), a marginally significant gain for cardiovascular death (relative risk 0.76, 95% confidence interval 0.58 to 0.99; NNT 112 for three years) and no effect at all on total mortality.¹² In contrast, a male smoker aged 50 with a systolic blood pressure of 145 mm Hg and a total cholesterol of 6.5 mmol/l is at low risk on SCORE.

A better way of using risk tables would be to compare the risk of an individual with the minimal risk of people of the same sex and age. Treatment should be considered when he or she has, for example, three times that minimal risk for his or her age and sex. This will prevent overtreatment of elderly people whose high risk is related to age and undertreatment of younger people who are at high risk. For our two examples the treatment options would be totally different.

All attempts to make risk tables more accurate, as done by Hippisley-Cox and colleagues in the QRISK2 algorithm,¹ are necessary and should be welcomed. However, this is not the key problem. We have to fundamentally rethink how to use risk tables when making treatment decisions in practice, taking into consideration the medicalisation of healthy older people and the correct use of drugs.