Assessing heart disease risk in primary care

Distilling evidence from randomised controlled trials and observational studies into valid and usable guidelines is not easy. Despite unequivocal evidence that lowering cholesterol concentrations reduces mortality from coronary heart disease,¹ producing guidelines on prevention that meet with universal agreement has proved difficult.² Four articles in this week’s issue illuminate the difficulties.

Unwin et al show that the application of different cholesterol guidelines leads to considerable variations in decisions to screen and to treat when applied to a representative population (p 1125).³ This is not surprising when the content and recommendations of the cholesterol guidelines are studied (see their table 1). A previous study in the United States showed a similar magnitude of disagreement between older and newer versions of the US guidelines, Canadian guidelines, and a coronary risk model derived from Framingham data.⁴ In the US study the Framingham model proved to be the most accurate method to predict future coronary heart disease mortality. On this basis it is comforting to see that two of the guidelines analysed by Unwin et al—the Sheffield and European guidelines—base their recommendations on an explicit calculation of absolute risk based on Framingham data. Indeed another set of guidelines from New Zealand also adopts a similar absolute risk approach.⁵

A consequence of discrepancy in guideline production is that the application of research evidence becomes inconsistent and may result in substantial variation when applied to individual patients⁶—as illustrated this week by Baxter et al (p 1134).⁷ They show that the prescribing of lipid lowering drugs has increased substantially since research evidence confirming the benefit of lipid lowering drugs has become available. What is most striking, however, is the substantial variation in the prescribing of lipid lowering drugs between general practices, with a 56-fold variation in nearly 500 general practices. Caution is always required in interpreting an ecological study such as this: as the authors acknowledge, many different factors may be responsible for such variation. However, the current controversy over the interpretation of cholesterol guidelines is unlikely to have helped general practitioners and their patients decide on the optimum course of action in managing a raised cholesterol concentration.

A recent Effective Health Care Bulletin has gone some way in resolving these difficulties by putting the issue of what to do about raised cholesterol into the context of the overall management of coronary heart disease.¹ When considering primary prevention it should be remembered that cholesterol concentration alone is a poor predictor of absolute risk of coronary heart disease. Absolute risk is determined by multiple factors, including age, sex, smoking status, blood pressure, and presence of diabetes and left ventricular hypertrophy.¹ As the relative benefit of cholesterol lowering by means of drug treatment is constant at about 30%, absolute benefit is determined by absolute risk.¹ Unfortunately, accurate assessment of the absolute risk of coronary heart disease in primary care appears to be poor. Doctors understand the relative importance of specific risk factors, but they tend to overestimate absolute risk in individual patients. This results in an overestimate of the absolute benefits of treatment, including the benefits accruing from a reduction in cholesterol concentration.⁸

Because of the difficulties of incorporating multiple risk factors into an accurate estimate of absolute risk, the starting point of primary prevention in any individual is an estimate of that individual’s absolute risk. This can be accomplished by looking up risk tables or using a computer program.⁵,⁹ One of the main computer suppliers to UK general practices now has a protocol that calculates absolute risk according to the Framingham risk equation.

Where guidelines differ is in the absolute risk level at which treatment should be started (annual coronary heart disease event rate: European 2%, Sheffield 3%). This is a matter of policy not an argument about evidence.¹ The recent Effective Health Care Bulletin endorses the approach taken by the authors of the Sheffield risk table. This is based on clear quantification of the cost effectiveness of treating individuals with an annual coronary heart disease risk of 3%.¹ The absolute level at which to start treatment has been subject to considerable disagreement,² but lowering the absolute risk threshold to 1.5% is likely to increase the percentage of people aged 35-69 who would be candidates for treatment from 3.4% to 20%.² A qualitative study in this week’s issue suggests that until consensus concerning the absolute risk level at which to start treatment is reached, general practitioners will continue to be uncertain about the role of lipid lowering drugs, particularly in patients at lower risk of cardiovascular disease (p 1130).¹⁰

There is no evidence to show that unselective population screening has any benefit; indeed the potential harms in terms of labelling and subsequent illness behaviour are poorly quantified and may be underestimated.¹ Secondary prevention of coronary heart disease requires lowering the overall risk by means of multiple risk factor intervention in the first instance: cholesterol should be lowered by drugs mainly because of the poor performance of lipid lowering diets in community settings.¹,¹¹

Finally, a paper by Pringle illustrates that implementing evidence based recommendations on preventing coronary heart disease at a practice level has substantial implications in terms of resources and opportunity costs (p 1120).¹² Gains in life expectancy for the practice population as a whole from risk factor modifications for coronary heart disease are modest but may be substantial in some individuals.¹³,¹⁴ In future it may be quantification of patients’ values about the trade off between lifelong treatment and prevention of coronary heart disease that will help when deciding on treatment in individual patients.⁶

General practice pp 1120-35