The Relation between Blood Pressure and Mortality Due to Coronary Heart Disease among Men in Different Parts of the World

Study Populations

Between 1958 and 1964, 12,761 men 40 to 59 years old who resided in seven countries were enrolled in the study.⁷ A total of 16 cohorts were included in the United States, Finland (eastern and western), the Netherlands (Zutphen), Italy (Rome, Crevalcore, and Montegiorgio), Greece (Crete and Corfu), the former Yugoslavia (Dalmatia, Slavonia, Zrenjanin, Velika Krsna, and Belgrade), and Japan (Tanushimaru and Ushibuka). In the United States and Rome, railroad workers were recruited. In the former Yugoslavia, workers from a large cooperative in Zrenjanin and professors from the University of Belgrade were invited. In the town of Zutphen, the Netherlands, for every nine men, four were invited to participate. In the remaining 11 cohorts (all rural), all men 40 to 59 years old who were listed in official registries were invited. Overall, the participation rate was greater than 90 percent, with several cohorts reaching participation rates of almost 100 percent.

To increase the power of the statistical analyses, the cohorts were pooled into six populations: the United States, northern Europe (eastern and western Finland and Zutphen), Mediterranean southern Europe (Montegiorgio, Crete, Corfu, and Dalmatia), inland southern Europe (Rome, Crevalcore, Slavonia, and Belgrade), rural Serbia (Zrenjanin and Velika Krsna), and Japan (Tanushimaru and Ushibuka).⁸ The criteria for pooling were similarities among cohorts in rates of mortality from CHD and similarities among cohorts in culture (such as dietary patterns) and geographic features.

Measurement of Blood Pressure and Clinical Assessment

In all 16 cohorts, major cardiovascular risk factors were measured according to standardized methods at enrollment, after 5 years (except in Japan), and after 10 years (except in the United States). Details of the methods used have been described previously.^7,9 Blood pressure was measured by a trained physician using a calibrated mercury sphygmomanometer on the right arm, with the subject in the supine position, at the end of the physical examination, according to the method later described in the World Health Organization (WHO) manual Cardiovascular Survey Methods. ¹⁰ Readings were taken to the nearest 2 mm Hg. The mean of two measurements, taken one minute apart, was computed for both the systolic and the diastolic blood pressure; for the diastolic pressure, the fifth-phase Korotkoff sound was assessed. Hypertension was defined as a systolic blood pressure of 160 mm Hg or higher, a diastolic blood pressure of 95 mm Hg or higher, or both. During the base-line period (1958 to 1964), medications designed to lower blood pressure were rarely prescribed in any of the seven countries, and therefore use of medication was not included in the definition of hypertension.

Nonfasting blood samples were drawn and serum total cholesterol levels were measured in standardized fashion at all laboratories according to the Abell–Kendall method, as modified by Keys et al.⁷ Current cigarette smoking was identified by a positive response on a standardized questionnaire. CHD at enrollment was defined as the presence of definite or possible myocardial infarction, according to predefined clinical and electrocardiographic criteria; definite angina pectoris, according to responses on the WHO questionnaire¹⁰; or chronic CHD manifested as heart failure or chronic arrhythmia, according to predefined clinical criteria.^9,11 When the Seven Countries Study began, it was not standard practice in clinical research to ask participants for written informed consent or to ask for approval from medical ethics committees.

Assessment of Mortality during Follow-up

To assess mortality in the study populations, all 12,761 subjects were followed for 25 years; 56 men (0.4 percent) were lost to follow-up. The underlying cause of death was coded by a single reviewer according to the criteria of the WHO International Classification of Diseases, 8th Revision (ICD-8).¹² The reviewer who coded the cause of death was blinded with respect to the subjects' cardiovascular risk factors. The final cause of death was established on the basis of information from the official death certificate (without other information in not more than 15 percent of all cases), from medical and hospital records, and from relatives of the person deceased or other witnesses and with use of a list of predefined criteria prepared by the primary investigators. In cases in which multiple causes of death were possible, priority was given to violent death, followed by cancer in an advanced stage, CHD, and stroke. The end point of the study was death during the 25 years of follow-up, with the primary cause established as CHD (ICD-8 codes 410 to 414), or sudden death from cardiac causes (ICD-8 code 795) when a coronary origin was mentioned.

Statistical Analysis

Among the 12,705 subjects with complete follow-up data, 246 (1.9 percent) had CHD at enrollment and for 16 (0.1 percent) data on CHD at enrollment were missing; these subjects were excluded from the analyses. We also excluded 412 subjects (3.2 percent) for whom data on covariates were missing, leaving 12,031 subjects in the analysis. For each population, the age-standardized 25-year rate of death due to CHD was computed by the direct standardization method, with use of the total study population as the reference population. In addition, for each population, the 25-year mortality from CHD, adjusted for age (in years), total cholesterol level (in millimoles per liter), and current cigarette-smoking status (no or yes), was computed per quartile of usual systolic blood pressure and usual diastolic blood pressure (with usual pressures calculated as described below). To do so, we first performed regression analyses for mortality due to CHD to obtain population-specific and quartile-specific regression coefficients for the three covariates (age, total cholesterol, and current smoking status). With these regression coefficients, we estimated multivariate-adjusted, population-specific mortality from CHD for each blood-pressure quartile, given the assumption that the mean level of the covariates for each population-specific quartile was equal to the mean level of the covariates for the total study population.

Cox proportional-hazards analysis, with the cohort as a stratification variable, was performed to estimate relative risks (SAS statistical software, version 6.12, PHREG procedure; SAS, Cary, N.C.). Relative risks of death from CHD were estimated by including either systolic blood pressure (in increments of 10 mm Hg) or diastolic blood pressure (in increments of 5 mm Hg) as a continuous variable in the model. Relative risks of death from CHD were also estimated with respect to the presence or absence of hypertension. In the multivariate analyses, adjustment was made for age and cohort as well as for age, cohort, total cholesterol concentration, and current cigarette-smoking status. To examine whether the relative risks differed among populations with different absolute risks of death due to CHD, we first created an ordinal population variable scored from 1 to 6, where 1 represented the population with the lowest age-standardized 25-year mortality due to CHD and 6 the population with the highest mortality. We subsequently tested for a significant interaction between this ordinal population variable and the blood-pressure variables (where P values of <0.1 by the likelihood-ratio test indicated significance, with one degree of freedom).

Short-term variations in blood-pressure values in individual subjects, resulting from imperfections in measurement or true biologic variability, bias the relation between usual blood pressure and mortality from CHD.^1,6,13 We corrected for this bias in two steps. First, for each subject, the usual, or average, blood pressures during the first five years of follow-up were estimated from a linear regression model, given the values obtained at enrollment and at the five-year follow-up for systolic and diastolic blood pressures, body-mass index, and cholesterol level.¹⁴ For each subject, the presence or absence of hypertension was then reassessed according to these estimates of usual blood pressures over the first five years. Second, the estimates of usual blood pressures and the new hypertension variable were analyzed in a Cox survival model to estimate regression coefficients for systolic and diastolic blood pressures and hypertension, with adjustment for within-subject variability. To examine the effect of within-person variability in blood pressure on estimated regression coefficients for systolic and diastolic blood pressures, we divided the adjusted regression coefficients by the unadjusted regression coefficients from the survival analysis to obtain population-specific adjustment factors.

Table 1. Table 1. Base-Line Characteristics of Men in the Seven Countries Study and Age-Standardized 25-Year Mortality from Coronary Heart Disease.

The base-line characteristics of each of the six study populations are shown in Table 1. The average systolic blood pressure at base line ranged from 132.5 mm Hg in Serbia to 143.7 mm Hg in northern Europe. The average diastolic blood pressure at base line ranged from 75.7 mm Hg in Japan to 86.6 mm Hg in both northern Europe and inland southern Europe. The proportion of men with hypertension was lowest in Serbia and Japan (15.7 percent and 16.2 percent, respectively) and highest in northern Europe (29.8 percent). The age-standardized 25-year rate of death due to CHD was low in Japan and Mediterranean southern Europe, intermediate in inland southern Europe and Serbia, and high in the United States and northern Europe.

Figure 1. Figure 1. Mortality Due to Coronary Heart Disease per Quartile of Usual Systolic Blood Pressure.

Values shown are 25-year rates of death due to coronary heart disease (CHD), adjusted for age, serum total cholesterol level, and cigarette-smoking status. The absolute risk of death at a given level of usual systolic blood pressure varied greatly among the populations.

Figure 2. Figure 2. Mortality Due to Coronary Heart Disease per Quartile of Usual Diastolic Blood Pressure.

Values shown are 25-year rates of death due to coronary heart disease (CHD), adjusted for age, serum total cholesterol level, and cigarette-smoking status. The absolute risk of death at a given level of usual diastolic blood pressure varied greatly among the populations.

In Figure 1, 25-year rates of death from CHD, adjusted for age, serum total cholesterol, and current smoking status, are plotted against the mean level of usual systolic blood pressure within the quartile. This plot shows that the absolute risk of death at a given value for usual systolic blood pressure varied strongly among the populations. For a usual systolic blood pressure of about 140 mm Hg, mortality varied by a factor of more than three, from approximately 20 per 10,000 person-years in Japan and Mediterranean southern Europe, to approximately 70 per 10,000 person-years in northern Europe and the United States. A similar pattern of variation in the absolute risk was observed for a usual diastolic blood pressure of about 85 mm Hg (Figure 2).

Table 2. Table 2. Age- and Cohort-Adjusted and Multivariate-Adjusted Relative Risks of Death from Coronary Heart Disease for Given Increments in Blood Pressure, before and after Adjustment for Within-Subject Blood-Pressure Variability.

For an increase of 10 mm Hg in systolic blood pressure, the multivariate-adjusted relative risk of death from CHD ranged from 1.09 in inland southern Europe to 1.25 in Serbia and Japan (Table 2). The multivariate-adjusted relative risk for all the populations combined was 1.17 before adjustment for within-subject variability in blood pressure and 1.28 after adjustment. For an increase of 5 mm Hg in diastolic blood pressure, the relative risk of death ranged from 1.06 in inland southern Europe to 1.19 in Mediterranean southern Europe, with a relative risk for the total population of 1.13 before adjustment for within-subject variation in blood pressure and 1.28 after adjustment. No significant differences were observed among the populations with respect to the relative risk of death from CHD over the 25-year period for these increments in blood pressure (P>0.1 by the likelihood-ratio test for the interaction between the blood-pressure variables and the ordinal population variable).

Table 3. Table 3. Multivariate-Adjusted Relative Risks of Death from Coronary Heart Disease Associated with the Presence of Hypertension.

The absolute risk of death from CHD that was associated with hypertension was clearly different among the six populations (Table 3). Among subjects with hypertension, the age-standardized 25-year mortality varied by a factor of nearly four, from 44 per 10,000 person-years in Japan and Mediterranean southern Europe to 153 per 10,000 person-years in northern Europe. Hypertension was a significant risk factor for death from CHD in all the populations: the relative risk before adjustment for within-subject variation ranged from 1.33 in inland southern Europe to 2.80 in Japan, and the overall unadjusted relative risk for hypertension was 1.77. When adjustment for within-subject variation was made by using usual values for systolic and diastolic blood pressure over the first five-year period, instead of single base-line values, to classify subjects as having or not having hypertension, the overall relative risk associated with hypertension became 2.13. No significant differences were observed among the populations with respect to the relative risk of death from CHD that was associated with hypertension (P>0.1 by the likelihood-ratio test for the interaction between the hypertension variable and the ordinal population variable).

We observed that the relative risk of death due to CHD in association with given increments in systolic and diastolic blood pressure and the presence of hypertension over 25 years did not differ significantly among the six populations, but that the absolute risk of death at the same level of blood pressure varied substantially. This indicates that the relation between blood pressure and the relative risk of death from CHD over the long term did not differ among populations in which the absolute risk of mortality from CHD varied considerably. This finding is consistent with our a priori hypothesis and agrees with results of large observational studies that have been confined to subjects from single populations.^2-4,15,16 The narrow confidence limits around the estimated relative risks of death for given increments in blood pressure indicate that our study had sufficient power to detect important differences in relative risks.

In all 16 cohorts of the Seven Countries Study, standardized methods were used to measure blood pressure, other cardiovascular risk factors, and causes of death.⁷ This consistency allowed valid comparisons to be made among populations. At the time of the base-line assessments (1958 to 1964), information on the use of antihypertensive drugs was not collected. However, during this period, such drugs were rarely prescribed in any of the seven countries. Furthermore, the estimated relative risks of death due to CHD over the first 10-year period of the study were similar to those for the 25-year period (data not shown). This similarity suggests that the initiation of antihypertensive therapy during the later years of follow-up did not influence the observed relations.

In survival studies of the relation between blood pressure and mortality, the estimated relative risks of death due to CHD associated with an increase of 10 mm Hg in systolic blood pressure varied from 1.2 in a study in Bergen, Norway, to 1.4 in the Western Collaborative Group study.^1,3,4,17,18 For a single measurement of diastolic blood pressure, the estimated relative risk for an increment of 5 mm Hg varied from about 1.1 in the study in Bergen to 1.3 in a cohort from three European cities (Edinburgh, Budapest, and Prague).^2,4,19 Estimated relative risks adjusted for within-subject variability in blood pressure have also been published.^1,6,20 In most of these reports, blood pressure was studied only in relation to total mortality from cardiovascular causes and to mortality from all causes.^6,20 The combined results of nine prospective, observational studies demonstrated that an increase of 5 mm Hg in diastolic blood pressure was associated with a 20 to 25 percent higher rate of death from CHD over a 10-year period, after adjustment for within-subject variability.¹ The estimated relative risks in the present study are similar.

Previous studies did not evaluate the effect of variability in blood pressure within individual subjects on the strength of the relation between systolic and diastolic blood pressures and mortality due to CHD. We observed that the effect was larger for diastolic pressure than for systolic pressure: the relation became 110 percent stronger for diastolic pressure and 60 percent stronger for systolic pressure after adjustment for within-subject variability. This difference occurs because the ratio of within-subject variability to variability between subjects is greater for diastolic than for systolic blood pressure.^21,22 Consistent with this finding is the fact that diastolic blood pressure, assessed as the fifth-phase Korotkoff sound by the auscultatory method, is more difficult to measure than systolic blood pressure.²³

Hypertension was a significant risk factor for death from CHD over the 25-year period of our study. The pooled relative risk associated with hypertension, defined as a systolic blood pressure of 160 mm Hg or greater, a diastolic blood pressure of 95 mm Hg or greater, or both, was 1.77 before adjustment for individual variability in blood pressure and 2.13 after adjustment. When hypertension was defined as a systolic blood pressure of 140 mm Hg or greater, a diastolic blood pressure of 90 mm Hg or greater, or both, the result was an estimated relative risk of 1.5. In the Framingham Study, hypertension was defined according to the latter criteria and was associated with a relative risk of 2.0 for death from CHD.¹⁵ The lower relative risk associated with hypertension in our present study may be due to our longer follow-up period of 25 years.

Because of random within-subject variability in blood pressure, analysis of single measurements of blood pressure may lead to substantial overestimation of the prevalence of hypertension, resulting in underestimation of the associated risk of death.²⁴ In the present study, the overall prevalence of hypertension decreased from 24 percent to 13 percent after adjustment for within-subject variability in blood pressure, yielding a 30 percent stronger relation with mortality due to CHD.

We observed substantial heterogeneity among the populations in rates of death due to CHD at similar levels of blood pressure. In the Seven Countries Study, similar heterogeneity has been observed for serum cholesterol levels.⁸ These differences in the absolute risk of death from CHD at similar blood pressures and serum cholesterol levels cannot be explained by differences in age or smoking status, because the analysis of mortality included adjustments for these factors. Although genetic differences among populations in susceptibility to CHD may partially explain the observed differences in CHD-associated mortality, other factors should also be considered. Differences in nutritional factors may play an important part, because dietary patterns vary greatly among countries.⁸ As compared with the diets in northern Europe and the United States, the Mediterranean diet at base line contained less meat and fewer dairy products but more olive oil, fish, fruits, vegetables, and alcohol.²⁵ In addition to interventions targeted to classic risk factors for CHD, changes in diet may therefore be important for reducing mortality from CHD in northern Europe and the United States to near the rates in the Mediterranean region and Japan. This process can be illustrated by the decrease in mortality from CHD that occurred at the same time as decreases in major risk factors in Finland during the period from 1972 to 1992.²⁶ A substantial increase in the consumption of vegetables and fruit in Finland starting in the early 1970s also contributed to the decline in mortality from CHD.²⁷

The large difference between the risks of CHD in the United States and northern Europe and those in Japan and Mediterranean southern Europe at the same blood-pressure level may have important implications for the treatment of hypertension. Recently, a task force of European and other societies on the prevention of CHD in clinical practice recommended the use of the absolute risk of CHD, based on all the major CHD risk factors, as a criterion for starting drug treatment.²⁸ According to this criterion, healthy persons whose absolute multifactorial risk of CHD will exceed 20 percent over the next 10-year period, or whose risk will exceed 20 percent if projected to age 60, have a sufficiently high risk to justify the selective use of proven drug therapies. The results of the Seven Countries Study imply that at the same blood-pressure level this criterion will be met at lower blood pressures in the United States and northern Europe than in Japan and Mediterranean southern Europe. Of course, the decision to start drug treatment is not based solely on absolute risk. Other factors, such as the clinical history, age, and sex of the patient and the cost effectiveness of therapy, are also important.

In conclusion, the present study showed that among populations, the increase in the relative risk of death from CHD for a given increase in blood pressure is similar but that the absolute risk at a given blood-pressure value varies substantially. If the absolute risk of coronary heart disease is taken as a criterion for the use of antihypertensive therapy, this finding will have major implications for clinical practice in different parts of the world.