ccounting for CHD risk: How much do major risk factors explain?

Cardiology Review® Online, June 2004, Volume 21, Issue 6

From the Department of Primary Care and Population Sciences, Royal Free and University College Medical School, and the Department of Community Health Sciences, St. George’s Hospital Medical School, London, United Kingdom

Despite impressive decreases in the incidence of coronary heart disease (CHD) over the last half century, it remains the single most common cause of death in Western societies and is an increasingly important problem worldwide.1,2 Observational cohort studies performed over the past 50 years have attempted to determine the factors that predispose certain individuals to CHD. These studies have conclusively demonstrated that cigarette smoking and elevated levels of serum cholesterol and blood pressure are three of the most important factors for developing CHD.3-5 Clinical trials have shown that reducing cholesterol and blood pressure reduces CHD risk.6,7 The strength of the relationship between these risk factors and CHD, however, has often been underestimated in prospective studies because risk factors were only recorded at a single baseline assessment. Baseline levels of cholesterol and blood pressure do not reflect true levels of the study cohort over the exposure period. Observed differences between individuals at baseline tend to exaggerate true differences over time because of measurement errors, random short-term variations, and longer-term systematic changes in individuals. This phenomenon, often referred to as regression dilution bias,8 has helped fuel the widely held belief that these three coronary risk factors account for only one half of CHD cases.9 This claim is important because it has implications for the likely success of prevention policies aimed at these risk factors and the potential importance of new, perhaps unknown, risk factors. In this study, we prospectively analyzed the relationship between cholesterol, blood pressure, and cigarette smoking and CHD risk after correcting for the underestimation caused by regression dilution bias. By partitioning individuals into groups at high or low risk of CHD, we estimated the proportion of all CHD cases that could be explained by these factors (the population-attributable risk fraction) and examined how these estimates varied with the thresholds used to define the low-risk group.

Patients and methods

Between 1978 and 1980, 7,735 men aged 40 to 59 years from 24 socially and geographically representative towns across Great Britain were enrolled in the British Regional Heart Study (BRHS).10 Seated blood pressure was measured and a nonfasting blood sample was taken for measurement of blood lipid levels. Cigarette smoking status was ascertained by a nurse-administered questionnaire. Men from two towns (one with high and one with low CHD mortality) were reexamined after 16 and 20 years of follow-up. A questionnaire provided information about physician-diagnosed CHD and current cigarette smoking habits. On both occasions, blood pressure and blood lipid measurements were taken.

All men were followed up for total mortality and cardiovascular morbidity since the baseline assessment. Information on mortality was collected through the National Health Service central registers; fatal coronary events were defined as deaths with ischemic heart disease as the underlying cause, including sudden death of presumed cardiac origin. Nonfatal myocardial infarctions (MIs) were ascertained from general practitioner reports, which were supplemented by systematic biennial patient record reviews. Diagnoses were confirmed according to es-

tablished World Health Organization (WHO) criteria. The analyses, which used the combined end point of nonfatal MI and death from CHD, were restricted to individuals who had no symptoms or diagnosis of CHD at the baseline examination, defined as no physician diagnosis of MI or angina, no evidence of angina on a WHO (Rose) angina question-naire, and no history of severe chest pain requiring a consultation with

a physician.

We calculated the population-attributable risk fraction, which is a function of the estimated relative risk of disease between the two groups and the estimated prevalence of high-risk exposure, for a range of criteria used to define low-risk individuals. We used this to estimate the proportion of major CHD cases that could be explained by high cholesterol, high blood pressure, or cigarette smoking. This was performed through logistic regression prediction both before and after correcting the regression coefficients for regression dilution bias.11


Of the 7,735 men recruited into the BRHS, 6,513 (84%) had no baseline evidence of CHD and had complete risk factor data on blood pressure, serum total cholesterol, and cigarette smoking status. The table shows the observed baseline levels of these men and the estimated usual levels over the first decade of follow-up, which was obtained by applying the estimates of regression dilution bias taken from the 4-year repeated data to the observed baseline data.

Over the first 10 years of follow-up, 426 men (6.5%) had a major CHD event (fatal or nonfatal MI). Figure 1 shows the relationships between usual levels of serum total cholesterol and systolic blood pressure and major CHD, adjusted for age, cigarette smoking, and each other. The relationships are continuous and graded, with no apparent thresholds below which a lower level does not confer a lower risk (at least down to 120 mm Hg of systolic blood pressure and 5 mmol/L of total cholesterol). These relationships were not affected by adjustments for other coronary risk factors, including body mass index, physical inactivity, and diabetes.

Figure 2 displays estimates of the relative risk and the population-attributable risk fraction for high serum total cholesterol, high systolic blood pressure, and current cigarette smoking considered simultaneously, both before and after correcting for regression dilution bias. Using the 20th centiles of the usual risk factor levels to define the low-risk group (< 5.5 mmol/L and < 131 mm Hg, respectively), and after correction for regression dilution bias, high-risk individuals had an estimated risk of major CHD 5.7 times higher than that of low-risk individuals (95% confidence interval [CI], 3.9—7.7), and the population-attributable risk fraction was 82% (95% CI, 74%–87%). Using the 10th centiles of usual risk factor levels to define the low-risk group (< 5.0 mmol/L and < 124 mm Hg, respectively), the relative risk was 7.9 (95% CI, 4.9–11.0) after correction for regression dilution, and the population-attributable risk fraction was 87% (95% CI, 80%–91%).


Defining low-risk individuals as being in the lowest quintile of usual levels of serum total cholesterol and systolic blood pressure and as current nonsmokers, the population- attributable risk fraction was 82% after correction for regression dilution bias. Therefore, if all individuals had the average risk levels found

in those in the low-risk group, 82%

of major CHD events within the

following 10 years would have been avoided. If everyone had the average risk levels of those in the bottom tenths of these distributions, 87%

of major CHD events would have been avoided.

In our analyses, we deliberately presented relative risks and population-attributable risk fractions as “curves” or functions of the “high-risk” criteria used to calculate them to show how they vary depending on how these groups are defined. Separation of individuals into low- and high-risk groups based on blood cholesterol and blood pressure levels is arbitrary. Evidence strongly suggests that the relationships between these factors and CHD have no threshold and continue below the levels of the bottom tenths of our study population.3-5,12 Thus, the effects of cholesterol and blood pressure exposure on CHD risk are likely to be appreciable, even in our low-risk group, compared, for example, with the Japanese cohorts of the Seven Countries Study,13 whose mean serum total cholesterol levels and CHD mortality risk were markedly lower than those in our lowest risk group. As a result, reducing risk factors to levels below those found in our low-risk group, if possible, would be likely to increase the proportion of prevented CHD events still further.

Our estimates of the population-attributable risk fraction before correction for regression dilution bias (70% to 80%) are consistent with previous studies. In the Multiple Risk Factor Intervention Trial, which included more than 270,000 men aged 40 to 57 years, CHD death rates between those at low risk (defined as serum total cholesterol < 5.17 mmol/L, blood pressure ≤ 120/80 mm Hg, and no cigarette smoking) and all remaining individuals were compared for patients who were initially free from disease. After 16 years of follow-up, the CHD death rate observed in the low-risk group was 78% lower than that in the rest of the sample.14 Similar estimates were obtained for men and women in the Chicago Heart Association Detection Project in Industry study14 and for men in the National Cooperative Pooling Project.15 In the Whitehall I study of more than 17,000 middle-aged British civil servants, it was estimated that if the average CHD mortality rate in the whole population could have been reduced to that found among individuals who had never smoked cigarettes and who were in the lowest quintiles of blood cholesterol and blood pressure levels, then about two thirds of the CHD deaths would have been avoided9; however, none of these studies was able to take into account the effects of regression dilution bias, although these effects were discussed in some cases.


In contrast to the “only 50% claim,” it is likely that at least four fifths of first major CHD cases in middle-aged men in Great Britain may be attributed to the three strongest coronary risk factors. This estimate will vary, however, depending on how the low-risk group is defined. Research may one day identify other common and important factors that predispose individuals to CHD. Until then, control

of the three most established fac-

tors alone could greatly reduce the CHD epidemic.