Editorials

Guidelines on preventing cardiovascular disease in clinical practice

Absolute risk rulesbut raises the question of population screening

Papers pp   671 , 676 , 677 General practice pp   680 , 686 , 690 Education and debate pp   702 , 705 , 709

Ten years ago clinical recommendations on preventing cardiovascular disease focused primarily on managing individual risk factors, particularly raised blood pressure and cholesterol concentrations. Typically, separate guidelines were developed for each risk factor and treatment was recommended when that factor was above a specified level.¹ The recommendations were informed mainly by evidence from cohort studies showing increased relative risks of cardiovascular disease in people with raised levels of the risk factor² and by evidence from randomised controlled trials showing relative benefits from lowering the factor. ^{3 4}

Over the past decade we have witnessed a remarkable change from these recommendations based on relative risk to ones based on absolute riskthat is, incidence. If Geoffrey Rose, arguably the most influential cardiovascular disease epidemiologist ever, was living today, he would support this revolution, which echoes his 1991 advice that "All policy decisions should be based on absolute measures of risk; relative risk is strictly for researchers only."⁵ This week's BMJ brings together a range of papers relevant to this paradigm shift in cardiovascular risk management.

One of the most recent examples of guidelines based on absolute risk is the Joint British Recommendations on the Prevention of Coronary Heart Disease in Clinical Practice, published late last year⁶ and summarised in this issue (p 705).⁷ Taking the lead from the European societies of cardiology, atherosclerosis, and hypertension, which jointly published coronary heart disease prevention guidelines in 1998,⁸ the British Cardiac Society, British Hyperlipidaemia Association, and British Hypertension Society joined forces to develop the current British recommendations. These new clinical guidelines recommend that priority for treatment should be given to patients at high absolute risk of coronary heart disease, defined as the probability of developing coronary heart disease over a specified period, rather than "undue emphasis being placed on an individual risk factor." ^{6 7} The British Hypertension Society has separately published guidelines for managing hypertension which are based on the same principle.⁹

Using absolute risk to inform clinical decision making around cardiovascular disease prevention is no longer seriously questioned. Many clinical trials have shown that the relative benefits of lowering blood pressure or blood cholesterol are similar for all patients at all ages up to about 75-80 years. ^{3 10 11} This trial evidence, now based on over 30 000 patients randomised to statins or placebo and followed for over five years,¹¹ supersedes lower level evidence from cohort studies⁴ suggesting an age-related differential in relative risk reduction. The paper by Ramachandran et al (p 677) is an example of analyses based on low level cohort study evidence measuring lipid related risk,¹² which is not supported by more appropriate trial data measuring lipid lowering treatment benefit. Their conclusions that patients stand to gain more through treatment of their main risk factor, is therefore likely to be wrong and simply reflects the weighting their analyses give to younger people with raised levels of individual risk factors. The trial evidence clearly shows that the absolute benefits of treatment are directly proportional to the pretreatment risk, over a wide range of ages, risk factor levels, or pretreatment risk. ^{3 10 11 13} The absolute risk of cardiovascular disease is strongly influenced by the combination of risk factors present, particularly a history of cardiovascular disease, age, gender, diabetes, smoking, blood pressure, and blood lipid concentrations.¹⁴

For example, a 50 year old non-smoking woman with a blood pressure of 170/100 mm Hg, a total cholesterol of 6.0 mmol/l, and a high density lipoprotein cholesterol of 1.2 mmol/l has about a 6% chance of suffering a major cardiovascular event in the next five years,¹⁴ whereas a 60 year old male smoker, with the same blood pressure and same total blood cholesterol values but a high density lipoprotein cholesterol value of 1.0 mmol/l has about a 30% risk.¹⁴ With antihypertensive or lipid lowering drugs, both these patients could reduce their risk of cardiovascular disease by up to a third over the next five years. ^{10 11} Therefore the 50 year old woman's absolute risk over the next five years could fall from about 6% to 4% (a 2% absolute gain), while the 60 year old man's could fall from about 30% to 20% (a 10% absolute gain). In other words about 50 such women but only 10 such men would require five years of treatment to prevent one cardiovascular event.

Absolute risk assessment charts are now included with many guidelines to enable clinicians (and their patients) rapidly to measure a patient's absolute risk of coronary heart disease or cardiovascular disease in the way described above. ^{8 15-18} Three examples of risk assessment charts appear in this issue (pp 705, 709, 672). ^{7 19 20} The current New Zealand and Joint British charts are similar in concept, both having been strongly influenced by the 1994 Joint European societies' charts.²¹ Both incorporate the same age categories and risk factors, with blood pressure and the total cholesterol: high density lipoprotein cholesterol ratio presented as continuous variables. While the New Zealand chart assesses five year risk of all cardiovascular disease in eight discrete categories, the Joint British chart assesses 10 year risk of coronary heart disease in three risk bands. The Sheffield tables, which were specifically designed to target lipid lowering therapy to patients at high absolute risk, simplify treatment decisions by using a truncated risk factor set and by only enabling estimation of risk above specified treatment cut off levels. Blood pressure levels are dichotomised, and only the most recent version of the tables include high density lipoprotein cholesterol. Like the Joint British charts, the Sheffield tables estimate the 10 year risk of coronary heart disease.

As most relevant interventions, such as smoking cessation, blood pressure lowering, and lipid lowering therapies, reduce the risk of both coronary heart disease and stroke, it would seem preferable to target cardiovascular disease rather than coronary heart disease risk. The recent British Hypertension Society guidelines acknowledged this⁹ but chose to recommend estimation of coronary heart disease risk to be consistent with the existing recommendations from the Joint British guidelines.⁶ As the two measures of risk are strongly correlated, multiplying coronary risk by 4/3 will give a reasonable estimate of cardiovascular risk.⁹

The original 1993 New Zealand charts, like the current UK charts, estimated 10 year risk, ^{22 23} but more recent NZ charts use five year risk ^{15 16} in response to feedback from clinicians that patients have difficulty personalising the longer risk period and because most treatment trials have run for about five years.^10-13 As well as risk, the New Zealand charts enable one to estimate the five year benefit of treatment, expressed as events prevented per 100 patients treated and as the number of patients needing treatment (NNT) for five years to prevent one event.

While each of the charts has gone through several iterations in response to informal feedback, two studies reported in this issue have formally evaluated aspects of the charts (pp 686, 690). ^{24 25} Isles et al randomly allocated doctors and nurses from 37 Scottish general practices to assess cardiovascular risk in 12 case histories, with one of the three risk charts.²⁴ Not unexpectedly, the New Zealand and Joint British charts performed equally well, and in terms of accuracy and personal preference they fared better than the Sheffield tables, despite the latter's simplicity. Montgomery et al assessed the impact of a computerised decision support system plus the New Zealand chart, the New Zealand chart alone, or usual care, on the management of hypertension in a cluster randomised trial in 27 Avon general practices.²⁵ Use of the chart alone was associated with a clinically significant reduction in blood pressure, due to more intensive drug treatment. Computerised support conferred no additional benefit, possibly because of the primitive nature of the computerised system. However, the study does show that clinicians are willing to use cardiovascular risk charts to support their management of hypertension in a busy primary care setting.

The risk charts discussed have all used risk prediction equations derived from the Framingham heart study.¹⁴ The equations were based on a 10 year follow up of about 5000 residents of Framingham, Massachusetts, from the late 1960s. Several studies, including one in this issue show that the Framingham equations can predict coronary risk with reasonable accuracy in white men and women in the United Kingdom (p 676). ^{26 27} However, little is known about the charts' predictive validity in high risk groups such as South Asians, Polynesians, or African Americans.

In an accompanying paper Robson et al thoughtfully discuss some of the outstanding questions for primary care arising from the assessment of absolute cardiovascular risk, in particular screening and risk thresholds (p 702).²⁸ Though assessment of absolute cardiovascular risk allows clinicians to target patients at high risk more effectively,²⁹ in essence it requires population screening of blood lipids, blood pressure, and the other risk factors included in the risk charts.

Wallis et al show that the new Sheffield tables can be used to accurately target patients requiring blood lipid measurements (p 671),²⁰ but almost everyone over the age of 45 years would require screening. Universal blood lipid screening (both total cholesterol and HDL cholesterol) above the age of about 45 years is the price we will have to pay to accurately identify patients at high cardiovascular risk. Selective screening will be relevant only in younger people.

The more important question raised by Robson et al is the appropriate risk threshold for treatment. The paper by our group (p 680)²⁹ illustrates one approach to choosing risk thresholds based on maximising events prevented without increasing the total numbers treated in a defined population. Answering the question about appropriate absolute risk treatment thresholds will be complex but is more relevant than asking what threshold blood pressure or blood cholesterol levels should be treated. Meanwhile, I endorse Robson et al's conclusion that we should initially focus on identifying the high risk patients whom we all agree should be treated.

Rodney Jackson, professor of epidemiology. 

Department of Community Health, University of Auckland, Private Bag 92019, Auckland, New Zealand