The accurate estimation of risk for future disease events is critical to the determination of the benefit–risk ratio and the most cost-effective use of preventive therapies (Box 1). This is particularly relevant for cardiovascular diseases (CVD), which are the leading cause of deaths in Australia (40% of total deaths), and in 1993–1994 accounted for the largest proportion (12%, or $3.9 billion) of total annual recurrent health expenditure.3 (This proportion is now almost certainly greater.) Expenditure on cardiovascular drugs under the Pharmaceutical Benefits Scheme totals $1.2 billion annually, $629 million of this on lipid-lowering drugs, especially statins.4 Accurate assessment of the likelihood of future events would optimise resource allocation by targeting patients at higher risk.5

In this context, the work of Simons et al, reported in this issue of the Journal (page 113),6 is very important. In their ongoing Dubbo Study (which commenced in 1988 and involved 2805 men and women aged 60 years and older when first assessed), the authors evaluated a risk function for coronary heart disease (CHD) prediction developed from a longitudinal cohort study in Framingham, Massachusetts — the Framingham Study. They also derived a risk function for future CVD events, including stroke as well as CHD, by modelling data from the Dubbo cohort.

The Framingham risk functions7-9 are widely used and form the basis of a New Zealand cardiovascular risk calculator,10 itself proposed as the absolute risk measurement tool in the recent lipid guidelines of the National Heart Foundation/Cardiac Society of Australia and New Zealand.11

The Framingham cohort consists primarily of white, middle-class individuals. The equation was derived from calculations based on age, sex, cigarette smoking status, diabetes status, and blood pressure, cholesterol and HDL cholesterol levels only. The Framingham measurements were also made some time ago before the dramatic increase in the prevalence of diabetes,12 and indeed Framingham included low numbers of people with diabetes.

In essence, the study by Simons et al determined the applicability of observations made in another time and another place to an Australian population. They showed that the Framingham equation accurately predicted overall 10-year incidence of "hard" CHD endpoints (myocardial infarction or coronary death). This supports previous validation work with the Framingham equation in the Busselton study.13 However, the Busselton study is now over 20 years old, while the Dubbo cohort included only older individuals.

Therefore, while these validation studies are important, it would be more relevant to derive predictive equations from data obtained from a representative and contemporary Australian cohort. This would acknowledge the variety of ethnic groups in Australia and the current mix of known and unknown risk factors. As the Framingham equation correctly predicts risk in only about 80% of cases,14 there is considerable interest in "novel" risk factors (eg, high sensitivity C-reactive protein) and techniques for imaging the arterial wall. Future research must examine the degree to which these elements might improve the ability to correctly identify those at risk.

With the shift of treatment guidelines from individual risk thresholds for treatment to decisions based on multivariable absolute risk, the logical extension of this is to estimate treatment efficacy or effectiveness in terms of absolute treatment benefit. For example, the benefits of cholesterol lowering in terms of improving average life expectancy have previously been estimated.15 This approach would provide more meaningful information to both patients and clinicians, as well as allowing the non-cardiovascular benefits of modifying risk factors such as tobacco smoking, physical inactivity, and unhealthy diet to be taken into account.

Data such as those from the Dubbo study have important implications for current Australian guidelines and practice (Box 2). Treatment decisions based on individual risk-factor thresholds are inadequate. The Dubbo study assesses the validity of the Framingham risk prediction equations in the elderly and provides an Australian risk equation for the same age group. Further work should be done to validate these equations in a wider contemporary population and to determine the extent to which new risk factors may improve assessment of risk. This should not, however, preclude swift measures to implement the use of absolute risk, as well as consideration of absolute treatment benefit, as ways of guiding treatment decisions in clinical practice.