Objective To estimate the burden of social inequalities in coronary heart disease (CHD) and to identify their major determinants in 15 European populations.
Methods The MORGAM (MOnica Risk, Genetics, Archiving and Monograph) study comprised 49 cohorts of middle-aged European adults free of CHD (110 928 individuals) recruited mostly in the mid-1980s and 1990s, with comparable assessment of baseline risk and follow-up procedures. We derived three educational classes accounting for birth cohorts and used regression-based inequality measures of absolute differences in CHD rates and HRs (ie, Relative Index of Inequality, RII) for the least versus the most educated individuals.
Results N=6522 first CHD events occurred during a median follow-up of 12 years. Educational class inequalities accounted for 343 and 170 additional CHD events per 100 000 person-years in the least educated men and women compared with the most educated, respectively. These figures corresponded to 48% and 71% of the average event rates in each gender group. Inequalities in CHD mortality were mainly driven by incidence in the Nordic countries, Scotland and Lithuania, and by 28-day case-fatality in the remaining central/South European populations. The pooled RIIs were 1.6 (95% CI 1.4 to 1.8) in men and 2.0 (1.7 to 2.4) in women, consistently across population. Risk factors accounted for a third of inequalities in CHD incidence; smoking was the major mediator in men, and High-Density-Lipoprotein (HDL) cholesterol in women.
Conclusions Social inequalities in CHD are still widespread in Europe. Since the major determinants of inequalities followed geographical and gender-specific patterns, European-level interventions should be tailored across different European regions.
Statistics from Altmetric.com
Social inequalities in health across Europe are a well documented issue accounting for more than 700 000 deaths per year and for an estimated 20% of the total European Union healthcare expenditure.1 International studies suggested that the magnitude of social inequalities varies across populations and by gender,2 ,3 but the lack of individual data limits the ability to investigate the underlying determinants for health inequalities and their heterogeneity across populations.
Attained education is considered a life-course socioeconomic indicator capturing an individual's knowledge-related resources.4 In a recent pooled analysis of prospective cohorts from nine European countries, educational inequalities in behavioural risk factors accounted for 29% and 34% of the excess risk of cardiovascular disease mortality observed among the least educated men and women compared with highly educated individuals, respectively.5 These findings cannot be directly extrapolated to inequalities in disease incidence because mortality rates are determined by case-fatality as well. Most prospective studies conducted in European populations have found an increased coronary heart disease (CHD) incidence among less educated individuals, although excess risks varied across populations.6–14 In these studies the prevalence of the ‘low education’ class ranged between 20%11 and 69%7 across the populations and from 50% to 24% across birth cohorts within the same population,14 reflecting differences in the schooling systems over place and time. Moreover, older subjects are more likely to be over-represented in lower educational classes,4 and most of the previous studies either provided gender-pooled findings6 ,8 ,9 ,13 or focused on one gender group only.7 ,10 ,11 Social disparities in disease are therefore of increasing concern but with limited comparative data to investigate their determinants across populations and gender groups. In this paper we aimed: (1) to assess the association between education and CHD mortality and incidence in different European populations, using a consistent definition of educational class across populations and birth cohort; (2) to disentangle inequalities in CHD mortality between incidence and 28-day case-fatality; and (3) to estimate the extent to which inequalities in CHD incidence can be explained by inequalities in distribution of major CHD risk factors.
Materials and methods
The MORGAM (MOnica Risk, Genetics, Archiving and Monograph) Project15 is a multinational collaborative study of prospective cohorts with standardised risk factors assessment and common follow-up procedures for major cardiovascular disease events. Details of MORGAM cohorts and data quality assessment for baseline and follow-up data are documented at http://www.thl.fi/publications/morgam. The present analysis is based on 49 cohorts from 14 populations recruited in 10 European countries, and 1 population from Russia. The baseline examinations took place mostly during the 1980s and early 1990s, although more recent cohorts are available for some populations (see table 1).
Definition of educational classes
Information on the number of years of schooling was collected at baseline (‘How many years have you spent at school or in full time study?’). The item's comparability across populations was high, and missing data prevalence was low.16 We derived three categories of education (high, intermediate and low) from population-specific, sex-specific and birth cohort-specific tertiles of the distribution of years of schooling.17 Furthermore, to take into account the distribution of education across populations, we used regression-based measures of inequalities.18 In each population and gender group, if a, b and c were the proportions of people in the low, intermediate and high education categories, respectively, the mean rank a/2, a+b/2 and a+b+c/2 was assigned to all subjects in the corresponding educational class. The rank variable was then used as the explanatory variable in regression models on the health measure of interest, to estimate the difference in health outcome between the most educated (rank=1) and the least educated (rank=0) subjects.19
Baseline CHD risk factors assessment
The baseline assessment of risk factors followed either the WHO-MONICA (Multinational MONItoring of trends and determinants in CArdiovascular disease) study protocol or MONICA-like procedures. Blood pressure was measured after 5 min rest while sitting and using either a standard or random zero sphygmomanometer, or an automated oscillometric device; when two consecutive measurements were available, the study variable for systolic blood pressure was the average value. Total cholesterol and High-Density-Lipoprotein (HDL) cholesterol were determined either in plasma or in blood sera. Daily cigarette smoking, use of antihypertensive treatment and the history of diabetes were derived from interviews or self-reported questionnaires; for the present analysis we combined former and never smokers as non-smokers. A positive history of CHD at baseline was defined as documented or self-reported history of myocardial infarction or unstable angina pectoris.
Follow-up procedures and end point definition
Death certificates with International Classification of Diseases, Ninth Revision (ICD-9) codes 410-414 or 798-799 (ICD-10 I21-I25, I46, R96, R98, R99) and hospital records with discharge diagnoses of ICD-9 410 (ICD-10 I21, I22) were investigated by all centres. Hospital records with discharge diagnoses of ICD-9 411 (ICD-10 I20) were investigated by all centres except Germany-Augsburg and Lithuania-Kaunas. Rehospitalisations within 28 days, or out-of-hospital deaths within 28 days, were considered as one event. Most of the centres further validated the event according to standard epidemiological definitions (see table 1).
The present analyses use two main end points, CHD mortality and incidence of major CHD events (first fatal and non-fatal myocardial infarction, non-fatal unstable angina pectoris, coronary death). As recurrent events were not registered in some centres, 28-day case-fatality was calculated on first events only, as the ratio between fatal and fatal plus non-fatal CHD events. Poland-Tarnobrzeg (no follow-up for non-fatal events), Poland-Warsaw and Russia (too few non-fatal events) contributed to the CHD mortality analysis only.
Of the available 113 485 men and women aged 35–64 years and free of previous CHD at baseline, we excluded 2557 (2.2%) due to missing data on years of schooling, leaving a final sample size of 110 928 individuals. For 4.5% of them one or more CHD risk factors were missing; we used multiple imputation techniques (SAS PROC MI, 10 imputed datasets; and PROC MIANALYZE) to maximise data availability.
We estimated the Slope Index of Inequality (SII) in CHD rates from Poisson regression models with the educational rank as the independent variable and adjusting for attained age during follow-up to mitigate the effect on rate estimates of different follow-up lengths across populations. The SII estimates the difference in CHD event rates (per 100 000 person-years) between the least and the most educated subjects, and it can be interpreted as the additional number of events attributable to educational inequalities.18 The SII was computed according to the formula in Mackenback et al (ref. 3, p. 2472), while 95% CIs were obtained through bootstrap (n=1000 samples with replacement; centile method). Educational class inequalities in incident mortality rates were disentangled as the sum of inequalities in total CHD incidence and in 28-day case-fatality, according to the method described in the online supplementary material.
As a relative measure of inequalities, we estimated the Relative Index of Inequality (RII) of CHD events from Cox regression models with attained age during follow-up as the time scale, the educational rank as independent variable and including dummy variables for study cohorts. The RII is interpretable as the HR for the least educated subject as compared with its most educated counterpart.18 Non-HDL cholesterol and HDL cholesterol, systolic blood pressure, antihypertensive treatment, smoking and diabetes, were added to the Cox models, one at a time at first and then all together. The % change in the age-adjusted RII for education due to a given risk factor was computed as:
All the analyses were stratified by sex and population. We also provided pooled estimates from fixed-effects models including a dummy variable for each population, and tested for homogeneity of educational inequalities by adding population×education interaction terms. All the analyses were carried out using SAS V.9.4.
During a median follow-up of 11.5 years (IQR: 8.9–18.6), 3244 CHD deaths and 6522 incident CHD events occurred among participants (table 1). Study populations include men and women aged 35–64 years, except France and Belfast (men aged 49–60 years only) and Denmark (birth cohorts of men and women aged 40 years, 50 years and 60 years at baseline). CHD mortality followed a well-known geographical pattern in men and women, with overall event rates being the lowest in France and southern Italy, and the highest in Scotland, East Europe and Russia (tables 2 and 3). Similarly, CHD incidence rates were 3.5 times and 6 times higher in Scotland than in southern Italy, in men (1005 vs 283 events per 100 000 person-years) and women (383 vs 60 events per 100 000 person-years), respectively.
Educational class inequalities in CHD mortality rates
The least educated men had higher CHD mortality rates (SII>0) than their most educated counterparts in all the investigated populations, and in particular in Russia-Novosibirsk (593 additional deaths per 100 000 person-years), Poland-Warsaw (471 deaths/100 000 person-years), the UK (293) and Finland (218; table 2). No clear geographical pattern emerged, as absolute inequalities were among the lowest in northern Sweden and in the rural population of Poland-Tarnobrzeg. In women (table 3), a lower CHD mortality rate for the least educated women was observed in the rural population of Poland-Tarnobrzeg (−106 deaths/100 000 person-years). In the remaining populations the SII was positive, and significantly greater than zero in the Nordic countries, in northern Italy and in Russia-Novosibirsk. Similar considerations hold when considering death from the first CHD event, in men and women.
Educational class inequalities in CHD incidence rates
CHD incidence rates and related SII values are reported in the last two columns of tables 2 and 3. The least educated men had higher CHD incidence rates than their most educated counterparts, with the pooled SII of 343 events per 100 000 person-years corresponding to almost half of the overall incidence rate (710.6 per 100 000 person-years). The pattern of inequalities in CHD incidence mostly reflected what was observed for CHD mortality, with the notable exceptions of Denmark and France (SII significantly greater than zero) and of northern Sweden.
In women, the SII for CHD incidence was statistically significant in most populations, including Germany-Augsburg and Scotland. The pooled SII value was 170 events per 100 000 person-years, corresponding to two-thirds of the overall incidence rate.
Breakdown of inequalities in CHD mortality into CHD incidence and 28-day case-fatality
Figure 1 (numbers are in online supplementary table S1) presents educational class inequalities in the rate of death from incident CHD event as the sum of inequalities in CHD incidence (in white) and 28-day case-fatality (in black). In men (left panel), inequalities in mortality were mostly attributable to event incidence in Scotland, Finland Lithuania, Denmark and northern Sweden. In the latter population, the 28-day case-fatality was lower in the least compared with the most educated men, resulting in a negative SII for mortality (−9.9 deaths/100 000 person-years, table 2). Conversely, 28-day case-fatality determined 36–73% of inequalities in mortality in Belfast and in central/south European populations (France, Germany, Italy). In women (right panel), inequalities in mortality rates were mostly attributable to incidence in 8 populations out of 10, the exceptions being northern Italy and northern Sweden. A lower case-fatality among less educated women was observed in southern Italy, Denmark and Scotland.
Explaining inequalities in CHD incidence with risk factors distribution
Educational inequalities in the distribution of risk factors are reported in the online supplementary table S2. RII in CHD incidence is reported in table 4 (men) and table 5 (women), while event rates and HRs in each educational class are shown as online supplementary table S3. The least educated men had a significant excess hazard of CHD events in Finland, Denmark, Belfast, Scotland and France (table 4), confirming the analysis of absolute inequalities reported above. The pooled RII estimate was 1.59 (95% CI 1.42 to 1.77), with no evidence of heterogeneity across populations (Wald χ2 test p value for homogeneity: 0.2). RIIs were attenuated after adjustment for non-HDL cholesterol, HDL cholesterol, systolic blood pressure, cigarette smoking and diabetes, remaining statistically significant in Finland and in Scotland; the estimated percentage of RII explained by these risk factors was 36%, but it varied across populations. Cigarette smoking and treated systolic blood pressure were major contributors, while non-HDL cholesterol had a relevant role in the Nordic countries populations only.
RIIs in women (table 5) ranged between 1.7 (95% CI 1.32 to 2.15) in Finland to 3.8 (1.3 to 11.1) in northern Italy, with no evidence of heterogeneity across populations (homogeneity test p value=0.4). The adjustment for traditional risk factors attenuated the pooled excess hazard by 36%, mostly due to HDL cholesterol (13% change) and systolic blood pressure (10%). Inequalities in smoking reduced RIIs in the Nordic countries and in Scotland, but not in central/South European populations and Lithuania.
In the 15 investigated populations of middle-aged European adults free of CHD, educational class inequalities accounted overall for 343 and 170 additional first CHD events per 100 000 person-years in the least educated men and women, respectively. These figures corresponded to 48% and 71% of the overall event rate in the two gender groups. The educational gradients were consistent across populations and gender groups and once the lower event rate has been taken into account, their magnitude was often larger in women than in men. The latter finding generalises previous observations from northern Italy12 and Denmark to other populations.14
In central/southern populations, a considerable proportion of inequalities in mortality from incident CHD (31–76%) could be avoided by eliminating inequalities in 28-day case-fatality. Conversely, in populations from the Nordic countries, Scotland and Lithuania more than 80% of inequalities in mortality were due to differential incidence. In northern Sweden, the proportion attributable to case-fatality was much higher for women than for men, possibly reflecting gender differences in the rate of admission to acute coronary care units.20 Inequalities in case-fatality have been mainly attributed to prehospital deaths,21–23 which reflect both events’ severity and delay in hospital presentation among less educated people.24 Registry data from Finland, Germany, Italy, France and Spain reported a long-term decline in 28-day case-fatality in men and women at a population level.25 Therefore, our findings suggest that the decline may have been uneven in some populations and gender groups, calling for specific interventions.
In our analysis of prospective cohorts, inequalities in the distribution of risk factors accounted for over a third of inequalities in CHD incidence in both sexes, consistent with the 39% change estimated by the INTERHEART case-control study.26 These estimates are interpretable as the expected reduction in inequality in CHD incidence due to interventions aimed at ‘levelling’ the risk factors’ distribution in the least educated to that of the most educated individuals.27 Prospective cohort studies have shown that a change in the distribution of risk factors had a larger impact on the observed cardiovascular disease mortality28 and CHD incidence29 in manual than in upper non-manual workers, providing a rationale for targeting risk factors in the lower social classes. The latest recommended actions to reduce CVD inequalities in Europe are mostly focused on improving access to healthy diet and physical activity, and reducing smoking and obesity rates in more disadvantaged people and areas.30 In the pathway linking education to CHD, smoking was a strong mediator in men, accounting for 20% of the excess hazard, but not in women from central/South and East Europe, where less educated women were less likely to smoke (see online supplementary table S2). The smoking advantage may have been lost in recent years, more notable in the youngest generations.31 ,32 Obesity13 ,14 and diet33 have also been found to be major mediators by recent prospective studies. In our populations, body mass index alone attenuated the pooled RII by 10% in men and 15% in women, but when added to the full model including systolic blood pressure and cholesterol it was no longer significant nor led to any further reduction in the RII (data not shown). Dietary advice has been found to be effective in reducing systolic blood pressure but not HDL cholesterol,34 the strongest mediator of inequalities in women. Since our findings suggested that the major determinants of social inequalities in CHD varied across populations and gender groups, they are well placed to inform European guidelines for CHD treatment and CVD prevention and how they could suggest more appropriate tailoring of interventions in lower socioeconomic strata across different European regions. In the UK the ASSIGN risk score35 includes a deprivation index to improve social equity in risk stratification, and the addition of education to the SCORE Project equation has been recently recommended in most European populations.36
We acknowledge several study limitations. France and Belfast cohorts were partially factory-based and we may have underestimated absolute inequalities in those populations. Risk factors were measured only once at baseline, leading to potential residual confounding when estimating the effect of risk factors on CHD inequalities, in particular for smoking (more educated subjects more likely to quit), non-HDL cholesterol (more educated subjects more likely to initiate statin treatment during follow-up) and systolic blood pressure (better control among the most educated subjects). In our populations former smokers constituted 49% of non-smokers in men and 18% in women. The average time since quitting was 12 years, with no educational gradient in men nor in women; the CHD excess risk attenuates within 5 years after cessation and continues to decline with time.37 Thus our definition of never-smokers is not likely to have substantially affected the estimate of the effect of smoking in our study. Most events were validated according to the standard MONICA or to other epidemiological definitions. Validation studies are available in some populations which used administrative sources to define the events, although results are not stratified by socioeconomic status.38 ,39 Participation rates were below 60% in two populations and ranged between 65% and 77% in the remaining ones (table 1). Non-participation affects the estimate of absolute inequalities in CHD rates,40 potentially introducing some heterogeneity in the comparison across populations. The SII and the RII measures assume a linear relationship between social rank and health,19 while some saturation effect could be present. We attributed social ranking based on educational class and not directly on years of schooling, and we do not have power to detect any departure from linearity.
Among the study strengths, we provided absolute and relative measures of educational class inequalities in CHD incidence in several European populations using prospective cohort studies with widely standardised measurement of risk factors and thorough end-point assessment. Comparable measurement of educational class inequalities in health between countries and over time is challenging and may introduce artificial heterogeneity.4 Our approach, taking into account birth cohort and the distribution of educational categories, mitigates the effects of differences in educational systems across countries and time periods, thus overcoming many of the problems.
To conclude, educational class inequalities in CHD mortality and incidence are still widespread in Europe, and their major determinants varied across populations and gender groups. Policies targeting risk factors in the most disadvantaged classes may reduce inequalities in CHD incidence by about a third, but interventions should be tailored in different European regions to effectively reduce the social gradient across all of Europe. Further research is needed to uncover and attribute the origin of the proportion of the social disadvantage in CHD incidence unaccounted for by traditional risk factors.
What is already known on this subject?
Low socioeconomic position has been found associated with higher cardiovascular disease and coronary heart disease (CHD) mortality rates in Europe.
However, no comparative estimates of the social gap in CHD incidence for different European regions and in men and women are available, so its major determinants are not known.
What might this study add?
Educational class inequalities in CHD incidence were largely consistent across populations and gender groups, and accounted for 48% and 71% of the overall event rate in men and women, respectively.
Twenty-eight-day case-fatality accounted for 31–76% of the educational class inequalities in CHD death rate in central and South European populations, while in the Nordic countries, Scotland and Lithuania the social gap in CHD incidence explained more than 80% of the mortality gradient.
Cigarette smoking was the major determinant of inequalities in CHD incidence in men, and High Density Lipoprotein cholesterol in women.
Inequalities in risk factors accounted for a third of the CHD educational class gradient, raising the need to further investigate the residual unexplained proportion.
How might this impact on clinical practice?
Social inequalities in CHD are still widespread in Europe, but with a North-South gradient in major determinants.
This study can inform European guidelines for CHD treatment and cardiovascular disease prevention which should tailor interventions in the lower socioeconomic strata appropriately across different European regions.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Contributors MMF and GV conceived the research and wrote the manuscript. GV conducted the statistical analyses, with the contribution of KK and LEC. KK directs the MORGAM Project and is the overall guarantor of the MORGAM data. All the authors actively contributed to the interpretation of the results and made critical revision of the manuscript drafts for important intellectual content.
Funding This work was supported by the MORGAM Project’s recent funding: European Community FP 7 projects ENGAGE (HEALTH-F4-2007-201413), CHANCES (HEALTH-F3-2010-242244) and BiomarCaRE (HEALTH-F2-2011-278913). These grants supported central coordination, workshops and part of the activities of the MORGAM Data Centre, at the National Institute for Health and Welfare (THL) in Helsinki, Finland. MORGAM Participating Centres are funded by regional and national governments, research councils, charities and other local sources.
Competing interests GV's work was supported by the Health Administration of the Regione Lombardia (grant n. 010804/2009) and by the European Community FP7 project BiomarCaRE (HEALTH-F2-2011-278913). KK received support from the European Union (HEALTH-F4-2007-201413; HEALTH-F3-2010-242244; HEALTH-F2-2011-278913). AP was supported by a grant from the Polish State Committee for Scientific Research. SV was supported by the Finnish Foundation for Cardiovascular Research.
Ethics approval Each MORGAM participating centre is responsible for ethical approval and patient consent, according to local rules at the time of study enrolment.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.