Despite a decline in cardiovascular disease (CVD) and ischaemic heart disease (IHD) mortality in Western countries, these conditions have risen greatly in Asian countries.1 In northern European countries and the US, higher socioeconomic position (SEP) is associated with lower IHD.2 However, the opposite has been reported in many developing countries3 as well as several southern European countries,4 especially among older men.4^–6 This contrasting pattern can be attributed to differences in the social distribution of major cardiovascular risk factors such as smoking,7 hypertension,8 and serum cholesterol.9 Based on the calculation of population attributable risk, several studies have shown the important role of these risk factors in IHD in Western and Asian countries.10^–13 Many studies from the West have examined the role of cardiovascular risk factors to explain the relationship between SEP and CVD/IHD mortality,14^–16 but often concluded that conventional risk factors did not explain most socioeconomic inequality.2 14 17 While those studies examined the relative ability of risk factors to explain socioeconomic differentials in health, inequality can also be viewed from an absolute perspective. Recently, Lynch and colleagues showed most of the absolute socioeconomic inequality in IHD could be attributed to conventional risk factors.18 Apart from this, no studies have been found that examined both the relative and absolute contribution of risk factors to explain socioeconomic health inequality.

South Korea has become a rapidly industrialised society, moving quickly through the epidemiological transition detected in Western countries during the 20th century. Along with industrialisation and an associated westernisation of lifestyle, CVD and IHD have emerged as major causes of death.19 20 Haemorrhagic stroke accounts for most stroke mortality in South Korea while IHD is relatively low.20 Unlike other Far East Asian regions, South Korea’s IHD mortality has shown an increasing pattern,19 20 growing more than fivefold among men and sevenfold among women from 1983 to 2004 (see fig 1). This suggests that Korea is currently undergoing a rise in epidemic IHD similar to that observed in Western countries in the early part of the 20th century. In Korea, a cross-over pattern of socioeconomic IHD mortality according to age was found.6 Considering the possible cross-over pattern by age in the social gradient of cardiovascular risk factors, age-specific analysis on the role of cardiovascular risk factors on SEP inequality in CVD may provide different policy implications for different age groups. However, age-specific analysis on the contribution of pathway variables to socioeconomic health inequality has rarely been performed. Prior South Korean studies have examined pathways to SEP inequality in mortality,21^–23 but no studies have been found that examined age-specific patterns in the role of pathway variables. Therefore, the purpose of this study was to examine age-specific patterns in the ability of major cardiovascular risk factors to explain relative and absolute socioeconomic inequalities in mortality from all causes, CVD, and IHD in South Korea.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Trends in age-standardised ischaemic heart disease mortality rate per 100 000 in South Korea, 1983–2004 (standardised to European standard population).

METHODS

Subjects

Data were obtained from the 1994 public servants’ health insurance beneficiary data, which included all South Korean government employees and teachers. Unique 13-digit personal identification numbers were linked to data on cause-specific mortality from the Korea National Statistical Office. Death certificate data after the 1990s are known to be complete.6 20 The proportion of deaths certified by physicians, the most important factor to improve the level of accuracy of cause-of-death information,20 was 90.3%. Because of the small number of female public servants, their young age and the small number of IHD deaths, women were not included in this study. Of the 692 645 men aged 30–64 (9.5% of total South Korean men aged 30–64), personal identification numbers of 0.7% (n = 4692) were invalid. A total of 10.7% (n = 74 541) did not participate in the 1994 health examination or had missing data on study variables. An additional 5.5% (n = 38 035) were excluded due to pre-existing disease. Of the remaining 575 377 men (83.1%), 16 998 died between January 1995 and December 2003. Causes of death were identified from death certificates per the International Classification of Disease, 10th Revision (ICD-10) codes.

Since the data used in this study have been collected routinely, participant consent was not specifically obtained. This study was approved by the Institutional Review Board of the Asan Medical Center, Seoul, South Korea.

Socioeconomic position

The SEP measure was level of national health insurance contribution. Since health insurance contributions in South Korea are imposed proportionally based on monthly salary, this measure directly reflects income level. A total of 53 levels were initially used for the determination of health insurance contribution. Since salary levels correlated well with participant age (r = 0.44, p<0.0001), we divided income into high and low groups at the nearest mid-point in each 1-year age stratum. Use of income tertile or quartile produced similar results.

Cardiovascular risk factors

All insured public servants are required to participate in a biennial health examination. Cigarette smoking status was based on respondent self-report. Blood pressure (BP), fasting serum glucose, and serum total cholesterol were measured by medical staff at 455 local hospitals. These four CVD risk factors were chosen because they are considered to be major established cardiovascular risk factors13 and their ability to explain socioeconomic health inequalities has been examined in many studies.14^–16 Blood pressure was measured in the seated position by a registered nurse or trained technician with a standard mercury sphygmomanometer or automatic manometer. For manual manometers, systolic and diastolic BP were measured as the first and fifth Korotkoff sounds, respectively. Total cholesterol and glucose were measured from fasting blood samples. Each examination site had internal and external quality control procedures directed by the Korean Association of Laboratory Quality Control.24

For this study, CVD risk factors were collapsed into groups. Cigarette smoking status was categorised into five groups: never smoked, ex-smoker, smoked ⩽10 cigarettes per day, smoked 11–20 cigarettes per day, and smoked >20 cigarettes per day. Blood pressure was classified as normal or pre-hypertension (systolic BP<140 mm Hg; diastolic BP<90 mm Hg), stage 1 hypertension (systolic BP 140–159 mm Hg or diastolic BP 90–99 mm Hg), and stage 2 hypertension (systolic BP⩾160 mm Hg or diastolic BP⩾100 mm Hg). Fasting serum glucose was dichotomised as <7.0 mmol/l and ⩾7.0 mmol/l. Finally, four categories were used for serum total cholesterol: <4.1 mmol/l, 4.1–5.1 mmol/l, 5.2–6.1 mmol/l, and ⩾6.2 mmol/l.

To calculate the population attributable risk (PAR) and risk difference (RD), we also categorised subject data into high-risk and low-risk groups. High-risk groups were defined as those with current smoking, hypertension, high serum cholesterol, and high fasting glucose.

Statistical analysis

Cox proportional hazard models were used to estimate age-specific relative risks (RR) and the 95% confidence intervals (CI) of mortality by income and cardiovascular risk factors. PARs were calculated with IRAP (Interactive Risk Attributable Program) version 2.2.25 The PAR for each risk factor was adjusted for age, income, and other cardiovascular risk factors. Age-adjusted percentages of participants in each risk factor group by income were calculated with age adjustment to 1-year age groups according to the direct method, with the distribution of total sample as the standard. The relative explanatory contribution of risk factors to mortality inequality, or relative explanatory ability, was measured by the percent reduction of excess RR for income level due to the inclusion of specific risk factors to the baseline of Cox’s model. Absolute explanatory ability was defined as the per cent reduction of excess RD between the whole study population and a low-risk segment free from risk factors.18 To estimate the absolute explanatory ability, we calculated the RD associated with exposure to one, two (smoking and BP) or any one of four risk factors (smoking, BP, glucose, and cholesterol) then calculated the per cent reduction in RDs. Finally, we presented the explanatory power of risk factors on the relative and absolute scales when one, two (smoking and BP) or all four risk factors were considered.

RESULTS

The study cohort contained 5 116 340 person-years of follow-up. Of 16 998 total deaths, CVD (n = 3065), cancer (n = 7422), and other causes (n = 6511) accounted for 18.0%, 43.7%, and 38.3% of deaths respectively. Death from IHD (n = 952) accounted for 5.6% of all-cause deaths and 31.1% of CVD deaths, while death from stroke (n = 1478) accounted for 48.2% of CVD deaths. Of stroke deaths, death from haemorrhagic stroke (46.3%; n = 684) was greater than that of ischaemic stroke (19.0%; n = 281). While the percentage of CVD deaths per total deaths increased with age, from 16.0% (753 of 4706) for ages 30–44 to 20.5% (1117 of 5446) for ages 55–64, the proportion of IHD to CVD deaths decreased with age, from 35.1% (264 of 753) for ages 35–44 to 28.2% (315 of 1117) for ages 55–64.

Differences in the prevalence of cardiovascular risk factors by age are also shown in table 1. Hypertension, high serum total cholesterol, and high serum glucose increased with age, while the proportion of current smokers decreased with age. The proportion of low-risk populations free from these four risk factors decreased with age from 27.2% for ages 30–44 to 20.4% for ages 55–64.

Table 2 presents age-specific patterns in age-adjusted (within age group) RRs of mortality from all causes, CVD, and IHD according to income and cardiovascular risk factors. Income differentials in mortality from all causes, CVD and IHD were observed. For IHD mortality, the magnitude of RR for the low-income group decreased with age and was not significant (RR 1.22, 95% CI 0.97 to 1.53) in men aged 55–64. While the four cardiovascular risk factors were predictive of CVD and IHD mortality, the size of RR varied with outcome and age. For stages 1 and 2 hypertension and high serum cholesterol, the RRs of mortality from CVD and IHD were generally greater than that from all causes of death. The magnitude of RR for hypertension was generally greater in CVD than in IHD. The magnitude of RR for stage 2 hypertension and high serum cholesterol generally decreased with age. This pattern was true for RRs of all-cause, CVD, and IHD mortality. Conversely, while RRs of CVD and IHD mortality for current smokers decreased with age, RRs for all-cause mortality increased with age.

Table 2 also shows the PAR associated with cardiovascular risk factors. Of the four cardiovascular risk factors, current smoking and hypertension showed relatively greater levels of PAR. For hypertension and high serum cholesterol, PAR for CVD and IHD was generally greater than that for all-cause mortality. The magnitude of PAR varied with age, decreasing for current smoking but increasing for hypertension and serum glucose. While for all causes PARs associated with any of the four cardiovascular risk factors were less than 40%, PARs for CVD and IHD were between 53.9% and 71.3%. When criteria for defining an individual as low-risk became stricter for smoking and BP (never smoked, BP<120/80 mm Hg) with other factors remaining the same, PARs for CVD and IHD increased to between 64.5% and 92.1%.

Table 3 presents the age-adjusted prevalence of cardiovascular risk factors by income and age group. Results show that socioeconomic differentials in risk factor prevalence varied with age and risk factor. Consistent inverse relations between income and current smoking were observed. Inverse relations between income and hypertension were also found among men aged 30–44 and 45–54, but not in men aged 55–64. This same pattern was also found for high serum glucose. While no relations between income and cholesterol were found among men aged 30–44, a positive relationship was detected among men aged 45–54 and 55–64.

Table 4 shows RRs by income and RDs between high and low income. The greatest reduction in RR and RD was found in the model adjusted for smoking (for RR) or when current smokers were removed from the whole population (for RD). This was generally true for all age groups and true for deaths from all causes, CVD, and IHD, with the exception of IHD in men aged 30–44 (for RD). Blood pressure provided the second greatest contribution to RR and RD reduction. After adjustment for all risk factors, significant RRs of mortality from all causes, CVD, and IHD persisted in all age groups, with the exception of CVD and IHD among men aged 55–64. However, due to a small number of deaths from IHD, RD for IHD had a wide 95% CI. Table 4 also presents percent changes in relative and absolute risk of mortality from all causes, CVD, and IHD. The magnitude of reduction in percent change in absolute risk was generally greater than that in RR. For example, while all four risk factors explained only 15.2% of excess RR of all-cause mortality for low income among men aged 30–44, the absolute excess risk was reduced by 48.3% when all four risk factors were removed from the whole population. This pattern was generally true for all age groups and risk factors. Cigarette smoking and BP made a relatively greater contribution to the reduction in both relative and absolute mortality risk for low income group. The explanatory power achieved by all four cardiovascular risk factors was similar to that achieved by smoking and BP. In addition, the explanatory ability measured by percent changes in RR and RD differed by age. For BP, the explanatory ability among men aged 30–44 was generally greater than that among men aged 55–64. This was true for deaths from all causes, CVD, and IHD. For serum total cholesterol, negative percent changes for RR of IHD mortality appeared among men aged 55–64 but these negative percent changes were not evident among men aged 35–44.

DISCUSSION

Relation of cardiovascular risk factors with mortality from all causes, CVD, and IHD

Results of this study showed that all four cardiovascular risk factors were significantly associated with risk of mortality from all causes, CVD, and IHD. For hypertension and serum cholesterol, RRs were greater in the association with CVD and IHD than with all-cause mortality. The magnitude of RR for hypertension was generally greater in CVD than in IHD. Considering the disproportionate contribution of BP to subtypes of stroke,26 this finding indicates that, for this study cohort, haemorrhagic stroke in CVD death surpassed ischaemic stroke.

We reported the influence of major cardiovascular risk factors on IHD according to age, an assessment that has rarely been done in previous studies. Consistent with prior studies, our study found that RRs of IHD mortality decreased with age for hypertension, serum cholesterol, and cigarette smoking.27^–30 Our study also indicated that hypertension and current smoking showed relatively greater levels of PAR for CVD and IHD. This finding is similar to the recent INTERHEART study12 and a prior South Korean study.24

Relation of SEP with risk factors and cause-specific mortality

In developing countries, increased IHD risk is often more pronounced in individuals with higher SEP.3 8 9 This contrasts with many northern European countries and the US, where SEP has been shown to be inversely associated with IHD.2 From epidemiological transition perspectives it has been suggested that, especially in men, the relationship between SEP and IHD has changed direction over time.5 31 32 This was true in our prior study.6 In this study, however, the reversal was not evident. This discrepancy may be due to population differences between the two studies with public servants in this study being relatively well-educated. According to our prior study,6 men aged 55–64 with high school education had greater mortality risk than those with college education, although men with middle or less education had lower mortality risk than men with college education.

Selective survival may also contribute to age-related reduction in RR of IHD mortality by SEP. Individuals with a higher SEP may survive into old age, thus exhibiting lifetime cumulative exposures to cardiovascular risk factors. Individuals with low SEP may die earlier, with only the healthiest surviving into old age.4 However, prior research has shown that selective survival cannot fully explain the age pattern.33 Since we excluded individuals with known diseases, this possibility would be diminished. Our results found changing associations of SEP with hypertension, serum glucose and cholesterol. The weak relation of SEP with IHD mortality among older public servants in our study might be attributable to positive or null associations of SEP with these risk factors.

Relative and absolute ability to explain inequality in cause-specific mortality

Results of our study showed that the explanatory ability of risk factors in absolute socioeconomic inequality in mortality was greater than that in relative socioeconomic inequality in mortality. The absolute excess risk of all-cause mortality was reduced by 48.3% when the four risk factors were removed from the whole population, while all four risk factors explained 15.2% of excess RR for low income among men aged 30–44. This was generally true for all causes, CVD, and IHD, and true for all three age groups and four risk factors considered. Smoking and BP were the leading contributors in explaining both relative and absolute inequality in mortality from all causes, CVD, and IHD. Although serum cholesterol was predictive of mortality, it showed a negligible effect on the explanation of relative and absolute mortality differentials. This was because cholesterol was not patterned by SEP and average levels of cholesterol in the Korean population are low by Western standards.

In addition, the explanatory ability of BP and cholesterol differed by age. For BP, the explanatory ability among men aged 35–44 was generally greater than among men aged 55–64. For serum total cholesterol, negative per cent changes for RR of IHD mortality in low-income groups were evident among men aged 55–64 but not among men aged 35–44. These differing patterns may be attributed to the changing relation of SEP with those risk factors according to age groups.

We used each public servant’s salary level relative to age as a measure for SEP due to correlation between salary and age. This measure can be seen as a relative SEP and may imply that people compare themselves to people of the same age according to their salary level. However, since the conventional CVD risk factors explained most of the absolute mortality difference in CVD and IHD, the role of the direct psychosocial pathway independent of indirect behavioural pathways would diminish. In addition, use of absolute income produced similar results to those presented here.

Limitations and strengths of the study

This study was not without limitations. First, all risk factors considered were based on single measures, resulting in limited reliability and regression dilution bias.34 This would underestimate the real contribution of those risk factors in determining mortality and explaining mortality inequality. Second, the number of IHD deaths used to calculate RDs in the low-risk segment of the population was relatively small. This resulted in unstable percent changes in absolute risk of mortality from IHD. Additionally, women were not included in this study.

However, it should be noted that this study was conducted in a country with low mean levels of serum cholesterol and subsequent low IHD mortality. The large number of study subjects with information about major cardiovascular risk factors made it possible to examine the age-specific contribution of cardiovascular risk factors to determine IHD mortality and to explain socioeconomic inequality in IHD mortality with an absolute as well as a relative perspective.