Article Text

Download PDFPDF

Roles of allostatic load, lifestyle and clinical risk factors in mediating the association between education and coronary heart disease risk in Europe
  1. Blánaid Hicks1,
  2. Giovanni Veronesi2,
  3. Marco M Ferrario2,
  4. Hannah Forrest2,
  5. Margaret Whitehead3,
  6. Finn Diderichsen4,
  7. Hugh Tunstall-Pedoe5,
  8. Kari Kuulasmaa6,
  9. Susana Sans7,
  10. Veikko Salomaa6,
  11. Barbara Thorand8,
  12. Annette Peters8,9,
  13. Stefan Soderberg10,
  14. Giancarlo Cesana11,
  15. Martin Bobak12,
  16. Licia Iacoviello2,
  17. Luigi Palmieri13,
  18. Tanja Zeller14,
  19. Stefan Blankenberg14,
  20. Frank Kee1
  21. On behalf of the MORGAM/BiomarCaRE consortium
  1. 1Centre for Public Health, Queen’s University Belfast, Belfast, UK
  2. 2Research Center in Epidemiology and Preventive Medicine, Department of Medicine and Surgery, University of Insubria, Varese, Italy
  3. 3Department of Public Health and Policy, Institute of Population Health and Sciences, University of Liverpool, Liverpool, UK
  4. 4Department of Public Health, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
  5. 5Cardiovascular Epidemiology Unit, Institute of Cardiovascular Research, University of Dundee, Dundee, UK
  6. 6Department of Public Health and Welfare, Finnish Institute for Health and Welfare, Helsinki, Finland
  7. 7Catalan Department of Health, Barcelona, Spain
  8. 8Institute of Epidemiology, Helmholtz Zentrum München, German Research Center for Environmental Health, Neuherberg, Germany
  9. 9German Centre for Cardiovascular Research (DZHK), Partner site Munich Heart Alliance, Munich, Germany
  10. 10Department of Public Health and Clinical Medicine, Umea Universitet, Umea, Sweden
  11. 11Research Centre on Public Health, Department of Medicine and Surgery, University of Milan–Bicocca, Monza, Italy
  12. 12Department of Epidemiology and Public Health, University College London, London, UK
  13. 13Department of Cardiovascular, Endocrine-Metabolic Diseases, and Ageing, National Institute of Health, Roma, Lazio, Italy
  14. 14Department of Cardiology, University Heart and Vascular Centre, Universitätsklinikum Hamburg-Eppendorf, Hamburg, Germany
  1. Correspondence to Dr Blánaid Hicks, Centre for Public Health, Queen's University Belfast, Belfast, Belfast, UK; B.Hicks{at}qub.ac.uk

Abstract

Background Previous studies have shown that differential exposure to lifestyle factors may mediate the association between education and coronary heart diseases (CHD). However, few studies have examined the potential roles of allostatic load (AL) or differential susceptibility.

Methods 25 310 men and 26 018 women aged 35–74 and CHD free at baseline were identified from 21 European cohorts and followed for a median of 10 years, to investigate the mediating role of AL, as well as of smoking, alcohol use and body mass index (BMI), on educational differences in CHD incidence, applying marginal structural models and three-way decomposition.

Results AL is a mediator of the association between educational status and CHD incidence, with the highest proportion mediated observed among women and largely attributable to differential exposure, (28% (95% CI 19% to 44%)), with 8% (95% CI 0% to 16%) attributable to differential susceptibility. The mediating effects of smoking, alcohol and BMI, compared with AL, were relatively small for both men and women.

Conclusion Overall, the educational inequalities in CHD incidence were partially mediated through differential exposure to AL. By contrast, the mediation of the educational gradient in CHD by investigated lifestyle risk factors was limited. As differential susceptibility in men was found to have a predominant role in the accumulation of AL in low educational classes, the investigation of AL-related risk factors is warranted.

  • cardiovascular diseases
  • education
  • health inequalities

Data availability statement

No additional data are available from each of the cohorts. Data may be made available to researchers upon reasonable request.

https://creativecommons.org/licenses/by/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Statistics from Altmetric.com

Request Permissions

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.

Introduction

The degree to which social, biological and behavioural risk factors explain socioeconomic inequalities in cardiovascular disease (CVD) remains uncertain. Two (nonmutually exclusive) mechanisms are hypothesised to play a role.1 First, risk factors for disease are unequally distributed across socioeconomic groups (differential exposure), accounting for approximately half of the inequalities observed, with large heterogeneity across populations.2 3 Second, differential susceptibility posits that the effects of risk factors on CVD may differ across socioeconomic groups.1

One early study aiming to distinguish between differential exposure and susceptibility, reported an increased susceptibility of manual workers to job strain and myocardial infarction, however, these were not investigated simultaneously.4 Others have explored these mechanisms simultaneously for observed socioeconomic status (SES) inequalities in CVD, reporting mixed findings, likely reflecting variable methodologies and heterogeneity across populations. Two studies by Nordahl et al have applied three-way decomposition models, one of which found a substantial contribution of smoking to educational gradients in CVD mortality.5 6 Hussein et al, using Oaxaca Blinder Decomposition, concluded differential exposure to a number of factors accounted for most of the inequality in CVD incidence, but differential susceptibility to neighbourhood socioeconomic conditions was observed.7

However, there remains a lack of insight into possible causal pathways invoked by differential susceptibility. Furthermore, analyses have not investigated the potential role of susceptibility to allostatic load (AL), a measure of the physiological ‘cost’ resulting from exposure to chronic stress.8 AL has been associated with coronary heart disease (CHD)9 and is proposed as a candidate pathway linking low education to CHD through a differential susceptibility mechanism.1 10 11 Specifically, it has been proposed that the relationship between SES and AL accumulation may be mediated by psychosocial and behavioural risk factors and in turn AL may act as a mediator between SES and CHD incidence.12 13 Thus, differential susceptibility may contribute to a greater accumulation of AL in individuals with lower education and a disproportionate effect of AL on CHD incidence. Indeed, we recently disentangled the contribution of lifestyle factors to the educational class gradient in AL in terms of differential exposure and susceptibility.14 Following on this work, based on the same 21 European population-based cohorts, we aim to examine this second pathway, disentangling the contribution of differential exposure and susceptibility to AL and other mediators in CHD educational gradients applying modern counterfactual mediation models.

Methods

Study population

We used data from the BiomarCaRE project, which includes population-based cohort studies harmonised in the MORGAM (MONICA Risk Genetics Archiving and Monograph) Project with stored serum/plasma.15 16 21 European cohorts (from six countries as outlined in online supplemental table S1) with harmonised data on education, lifestyle factors and markers required for the allostatic risk score were included. Individuals aged 35–74 years (Belfast (men aged 49–60) and Brianza and Catalonia (individuals aged up to 66 and 67 at baseline, respectively)) with no history of acute coronary events or stroke CVD at baseline and with complete data for calculation of the AL score were included. All participating studies received approval by local ethics review boards.

Supplemental material

Educational level

The number of years of schooling was collected at baseline and education levels were classified into three categories (high, intermediate and low) as population-specific, sex-specific and birth cohort-specific thirds of years of schooling.17

Allostatic load

AL comprises primary measures (stress hormones), intermediary outcomes (response of metabolic, cardiovascular and inflammation systems to stress), resulting in allostatic overload. Finally, chronic stress dysregulation leads to the manifestation of disease (eg, hypertension, diabetes and obesity).18 19 AL score was computed using eight selected biomarkers corresponding to three different physiological systems: inflammation (C reactive protein), metabolism (high-density lipoprotein (HDL) cholesterol, total cholesterol, triglycerides, blood glucose (HbA1C in MONICA/KORA Augsburg) and body mass index (BMI)); cardiovascular (systolic blood pressure and diastolic blood pressure). Measurement details for biomarkers are outlined in online supplemental table S2. Markers of the neuroendocrine system (eg, epinephrine and cortisol) were not available. AL score was calculated based on the sum of an individual’s Z-score for each of the eight individual biomarker components. Before standardisation, markers with a skewed distribution were log transformed. As HDL cholesterol is inversely associated with CHD, this was inverted. The Z-scores were derived from population-specific, sex-specific and fasting status-specific mean and SD values.

Secondary analyses calculated AL subscores for the three individual physiological systems by summing Z-scores for the relevant markers only.

Other covariates

Cigarette smoking, ascertained by interview or self-reported questionnaire, was categorised as a five-class variable as nonsmokers (never and former smokers) and current smokers, which were categorised into three levels of cigarettes/day (≤10, 11–20, >20). Daily alcohol intake (in grams) was converted to average drinks per day, with 12.5 g of alcohol considered a standard drink.20 Alcohol use was categorised as abstainers (0 drinks per day), 1–2, 3–4 and ≥5 drinks per day (0, 1–2, ≥3 drinks per day for women). BMI was categorised as normal (<24.9 kg/m2), overweight (25–29.9 kg/m2) and obese (>30 kg/m2).

Outcome

Hospitalisations for CHD and death were identified through linkage to national or regional death registries, hospital discharge records, population-based registries or via participants. Most centres used standard epidemiological criteria to define coronary events, while some relied on routine cause of death diagnoses and hospital discharge diagnoses. There was variation in codes used depending on local ICD-coding practices (ICD-9/10). The main endpoint for this study was the occurrence of CHD (first fatal and nonfatal myocardial infarction, unstable angina pectoris, coronary death or unclassifiable death).

Statistical analyses

59 065 participants had valid data on education and were free of previous CVD at recruitment. 8737 were excluded because of missing data for AL score markers; no follow-up information or missing data on lifestyle factors. Of 51 328 participants (87% of the original sample) were included, with no substantial differences in this percentage across educational classes (data not shown). Cohorts are described in online supplemental table S1.

To assess the mediatory role of AL score and behavioural factors (smoking, alcohol and BMI) on educational class differences in CHD incidence through differential exposure and differential susceptibility (accounting for measured confounding of age, sex and study population), we applied three-way decomposition. This method decomposes the difference in CHD outcome between two educational classes (ie, the ‘total effect’ of education) into the sum of three components: the pure direct effect (PDE), the pure indirect effect (PIE) and the mediated interaction (MI).21 To estimate these three components, we used marginal structural models by fitting sex-specific additive hazards regression models with age as the time scale.22 23 The additive hazards model allows for the estimation of the ‘total effect’ as the additional number of CHD events (per 100 000 person-years) in individuals in one education class as compared with the reference group (high education). The estimated average PDE can be interpreted as the additional number of CHD events in the low versus high educational class that is not mediated by the mediator; the estimated average PIE is the additional number of CHD events due to the different distribution of the mediator (indicating differential exposure); and the estimated average MI is the additional number of CHD events in the low class due to the interaction between education and the mediator on the outcome (differential susceptibility). The sum of the two components is the total proportion of inequalities mediated by the mediator.5 Marginal equation models are described in detail in online supplemental methods. We report coefficients (online supplemental tables S3 and S4) and weight distribution (online supplemental figures S1–S4) for the underlying propensity score models, to document the positivity assumption as well as the cumulative CHD event rates at fixed attained ages during follow-up, as a measure of goodness of fit for the additive hazard models (online supplemental table S5).

Analyses were repeated (1) by study population, (2) for AL subscores, (3) stratifying participants in the predysregulation and dysregulation phases (including elevated blood pressure (>140/90 mm Hg), type 2 diabetes or obesity). Finally, analyses were conducted to examine the conditional exchangeability assumption, investigating the potential impact of unmeasured confounding of the exposure–mediator relationship on differential exposure and differential susceptibility estimates (outlined in online supplemental methods). Analyses were conducted using R and SAS V.9.4 (SAS Institute).

Results

3031 participants were diagnosed with CHD, with a greater proportion in the low educational group (table 1). AL score was lowest among those with high educational status. Those with low education were more likely to be current smokers and smokers in the high education category had a lower smoking intensity. The proportion of alcohol abstainers was highest among low education groups, however, men in the low education group had higher daily alcohol intake. Obesity, elevated blood pressure and diabetes, was highest among those in the low educational class.

Table 1

Distribution of allostatic load scores, behavioural and anamnestic CVD risk factors and follow-up endpoints across the educational classes in men and women 35–74 years old, free of cardiovascular disease at baseline

CHD incidence was greater among both men and women in the low education group compared with the high education group (table 2). Overall, AL-mediated educational inequalities observed in CHD, with the highest overall proportion-mediated observed for AL among women (36%). Of the 142 additional CHD events per 100 000 person-years in women, 28% (95% CI 19% to 44%) were attributable to the PIE and 8% (95% CI 0% to 16%) to the MI (differential susceptibility). In men, the overall proportion mediated by AL was 19% (PIE (differential exposure) 16% (95% CI 11% to 23%); with MI (differential susceptibility) of 3% (95% CI 0% to 6%)).

Table 2

Rate difference in additional coronary heart disease events per 100 000 person-years by educational level (decomposition of total effect into direct, indirect and mediated interaction effects) in men and women

The mediating effect of smoking was stronger among men than women, accounting for the highest overall proportion of observed mediation in men (table 2). The total effect, comparing low to high education, was 250 CHD events per 100 000 person-years, of which 28% was mediated via smoking (PIE=23% (95% CI 16% to 34%); MI=5% (95% CI −1% to 10%). In contrast, alcohol use and BMI did not mediate the association between educational class and CHD incidence as strongly as smoking, with no evidence of a differential susceptibility effect. However, the effect of differential exposure to alcohol and BMI was larger in women than in men.

The contributions of the AL subscores to the educational gradient in CHD are reported in table 3. Inflammation contributes, though modestly, to the educational gradient in CHD both in terms of differential exposure (men 12% (95% CI 9% to 18%); women 11% (95% CI 7% to 18%)) and differential susceptibility (men 5% (95% CI 2% to 8%); women 9% (95% CI 4% to 14%)). The contribution via the metabolic system was largely attributable to differential exposure, with a stronger effect in women. Cardiovascular system markers contributed minimally. Results stratifying by the predysregulation and dysregulation stage were largely similar to the main analyses across groups (table 4). Analyses stratifying by age revealed evidence of differential susceptibility to AL among those aged 35–60 years (men 5%–95% CI 2% to 9%; women 15%–95% CI 5% to 24%), but not for those aged 60–85 years (online supplemental table S6).

Table 3

Rate difference in additional coronary heart disease events per 100 000 person-years by educational level (decomposition of total effect into direct, indirect and mediated interaction effects) for allostatic sub-scores as mediators in men and women

Table 4

Rate difference in additional coronary heart disease events per 100 000 person-years by educational level (decomposition of total effect into direct, indirect and mediated interaction effects) for allostatic load as mediators, separately for individuals in the pre-dysregulation and in the dysregulation phases

Figure 1 depicts additional CHD events by educational level by gender in each cohort. While largely consistent across sites, a number of negatively MIs were observed for men in Italy-Latina and Germany-MONICA/KORA Augsburg and women in Italy-Brianza and Northern Sweden.

Figure 1

Additional coronary heart disease events per 100 000 person-years due to the mediated interaction of allostatic load, by educational level in men (A) and women (B) by study population and overall estimate.

Online supplemental figures S5 and S6 show that the main findings were generally robust to an unmeasured confounder if it had an effect of comparable magnitude to age or neighbourhood deprivation.

Discussion

In this study, we found evidence that AL contributes to the educational inequalities in CHD incidence. We observed differential exposure and differential susceptibility to the effects of AL on CHD risk, however, a majority of the mediating effects of AL (and other behavioural factors) were through differential exposure. Differential susceptibility effects were modest, the main pathway for the latter appearing to be through susceptibility to inflammation.

It is fairly well accepted that education is causally associated with cardiovascular outcomes and all-cause mortality.24 25 For example, a recent Mendelian randomisation (MR) study found that 3.6 years of additional education reduced the ‘predisposition’ to CVD by about one third,24 however, the mechanisms are largely unknown.24 While a 1-SD longer education was also associated with a 35% lower odds of smoking and 0.17 kg/m2 lower BMI, we know that polygenes predict only a small proportion of the population variance in such behavioural traits. Indeed, Kaufman calls for caution of MR methods predominance when we know little about the mechanisms by which risk factors trigger disease.26

The notion of differential susceptibility aligns with the sufficient–component cause model27 and, thus, may be understood as conditional or a feature of causal interaction. The possibility that SES inequalities are generated by differential exposure has been aligned with mediation, insofar as we pose the counterfactual question of what CVD outcomes would be in the low SES group, had they the same risk factor distribution as the high SES group.1 The extent to which observed SES differences are driven by differential exposure or susceptibility is important, as optimum policy responses should vary according to the balance of the two mechanisms.

Previous studies largely used regression methods to investigate the mediating role of behavioural factors in the association between education and CVD.28–31 Few have aimed to disentangle the contribution of behavioural factors, in terms of differential exposure and susceptibility, reporting mixed findings, possibly reflecting differing methodologies. Hussein et al, applying Oaxaca Blinder Decomposition, observed that inequality in CVD incidence between high and low SES was largely attributable to differential exposure to diabetes, hypertension, social environment and neighbourhood socioeconomic conditions.7 Contributions via differential susceptibility for smoking and alcohol were negligible. Nordahl et al, adopting comparable methods to ours, found educational inequalities in CVD mortality were mediated considerably through behavioural factors, notably smoking (26% for men and 34% for women), with a significant effect via differential susceptibility, particularly for women (20%).5 In our study of CHD incidence, the effect mediated by smoking was similar for men but smaller for women (28% and 11%, respectively), and the proportion mediated via differential susceptibility was small (5%). However, Nordahl et al investigated cardiovascular mortality, focused on older people and disparate findings may reflect differences in the effect of smoking on educational inequalities in CVD across European populations.32

No previous study has investigated the contribution of differential exposure and susceptibility to AL in educational inequalities in CHD. AL was a mediator of the educational gradient in CHD, with mediation largely via differential exposure. In the same populations, we previously found that the educational gradient in AL in men was largely attributable to differential susceptibility to behavioural factors.14 In particular, being a never smoker or having moderate alcohol intake was less protective in terms of AL accumulation in less educated men compared with their more educated counterparts. In women, the educational gradient in AL remained largely unexplained by both differential exposure and susceptibility to the same behavioural factors. Taken together, our two companion analyses suggest the need to identify other factors causally linked to differential AL accumulation in lower social classes.

We observed some differential susceptibility to the effects of AL on CHD incidence, the main pathway for which was via inflammation. The importance of inflammation in CHD has been reported.33 Contrastingly, evidence suggests that cortisol, which modulates inflammation, may be associated with CHD.34 Unfortunately, markers of the hypothalamic–pituitary–adrenal axis were unavailable in this study. However, it is differential exposure rather than susceptibility to AL that explains more of its observed mediation of the educational inequalities in CHD incidence. So, if there is a policy implication, it may be that even if certain groups are more vulnerable to the accumulation of AL, in all likelihood we must act early in the life course to mitigate this and Marmot’s appeal to proportionate universalism does justice to the joint mechanisms revealed by our analysis.35

Despite this, we must also be mindful that the contribution of ‘traditional risk factors’ may vary across populations and may not explain all of the disparities observed.36 It is possible that in these middle-aged cohorts, it is already too late to mitigate the effects of differential susceptibility to AL, and the sensible policy response is to focus on minimising exposure—such as taxation, regulations and upstream targeting of the generative factors affecting other harmful behaviours. This coheres with evidence for intergenerational early life transmission of health behaviours and with accumulating evidence for the epigenetic embedding of early life stressors and their effects on adult disease risk.37 38 It is worthwhile noting that we did not control for certain factors that might moderate the consequences of AL such as marital stability and spousal education level, which can affect the consequence of stressors in adult life.39 Finally, it is salutatory to be reminded in a theoretical exposition by VanderWeele, built on a sufficient cause framework, that while statistical mediation implies mechanism, mechanisms may exist which are not tractable by current methods of statistical mediation.20

Strengths and limitations

This study included a large sample size, from 21 cohorts with long follow-up periods, providing adequate power to conduct mediation analyses, investigating both differential exposure and susceptibility.

However, we considered a number of mediators individually (not adjusting for other mediators) and the assumption that the mediators do not affect each other may be difficult to justify. It is likely they may be related, through direct pathways or common causes, affecting both occurrence and the effect of other mediators (differential exposure and susceptibility) (figure 2). Methods are only now being developed for multivariate and high dimensional cases.40 In particular, we recognise the possibility that AL accumulation itself may influence other lifestyle factors such as smoking. However, in this study, the median age of smoking initiation was 18 years and 17 years in women and men, respectively (data not shown), while AL accumulation was captured at baseline. Furthermore, a previous study found no evidence of a correlation between AL at age 9 and smoking at age 17, while smoking was a mediator for AL accumulation at age 17.41 Indeed, the estimation of direct and indirect effects requires strong assumptions, namely, no unmeasured confounding of the exposure–outcome, mediator–outcome and exposure–mediator relations, and no mediator–outcome confounder affected by exposure. While we could adjust for age and study population, there may be residual confounding due to unmeasured confounders or imprecisely measured covariates. Reassuringly, sensitivity analyses investigating unmeasured confounding revealed that an unmeasured confounder would need a strong correlation (stronger than age and neighbourhood deprivation) with both education and AL to affect our observed estimates (online supplemental figures S5 and S6). Cholesterol and blood pressure were included in our AL score. However, these may not only influence CHD via stress accumulation but also via other pathways. In addition, we were unable to include biomarkers from the hypothalamic–pituitary–adrenal axis or immune system in our AL as these were not captured consistently across all populations. It is of note that when compared with MR approaches, regression-based analyses of mediation somewhat underestimate the proportion mediated, reflecting the latter approach being more susceptible to measurement error.20 When measurement error of a mediator is differential across exposure levels, the interaction (differential susceptibility) is considered liable to underestimation.42

Figure 2

Directed acyclic graph describing the intertwined pathways of education on CHD through allostatic load, smoking and alcohol use. CHD,coronary heart disease.

Finally, AL and other mediators were measured at baseline, and for some cohorts, this was many years ago (1980s–1990s), thus the distribution of exposures has likely changed over time, particularly in the well educated. As we did not capture changes across time, this could lead to an underestimation of the mediated effect.43 Future studies, including repeated measures over longer follow-up periods, are warranted.

Conclusion

In this prospective cohort study, we found evidence for the effect of differential exposure to AL on CHD incidence. As differential susceptibility in men was found in our companion paper to have a predominant role in the accumulation of AL in low educational classes,14 further investigation of AL-related risk factors is needed. Meanwhile, any preventive action aiming to control factors linked to the disproportionate exposure to excess AL may help to reduce CHD morbidity, in particular among those with lower education.

What is already known on this subject

  • Differential exposure to behavioural and biological risk factors accounts for a limited proportion of social inequalities in cardiovascular diseases.

  • Allostatic load, a measure of the physiological ‘cost’ resulting from chronic stress, has been associated with socioeconomic gradients in coronary heart disease, but no studies have investigated the potential role of susceptibility to allostatic load.

  • We applied modern counterfactual mediation methods to examine the role of differential exposure and differential susceptibility to allostatic load and other lifestyle factors on educational gradients in coronary heart disease in Europe.

What this study adds

  • Allostatic load contributes to the educational inequalities observed in coronary heart disease incidence in Europe.

  • Overall, in men and women, allostatic load-mediated inequalities in coronary heart disease were largely driven by differential exposure (16% and 28%, respectively), while effects of differential susceptibility were modest (3% and 8%, respectively).

  • There was limited evidence of differential susceptibility to other mediators (smoking, body mass index and alcohol use).

Data availability statement

No additional data are available from each of the cohorts. Data may be made available to researchers upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

All participating studies adhered to the Declaration of Helsinki and are responsible for ethical approval and patient consent, according to local rules at the time of study enrolment. For different study populations, the list of approvals is the following: (study name, Ethics Committee name, approval ID): Northern Sweden, Research ethic Committee of Umea University, 2012-280-32M; FINRISK 1997, Ethics Committee at National Public Health Institute of Finland, 38/96; PRIME-Northern Ireland, Office for Research Ethics Committees Northern Ireland, 06/NIR02/107; MONICA/KORA Augsburg, Ethik-Kommission Bayerische Landesärztekammer, 05004; MONICA-Brianza, Comitato Etico Azienda Ospedaliera San Gerardo—Monza, 192/2005; MATISS Study (Latina), Comitato Etico Istituto Superiore di Sanità, PRE/96/06; Moli-Sani study, Comitato Etico Università Cattolica del Sacro Cuore—Roma, Prot.Pdc.P99 (A. 931/03-138-04)/CE/2004. MONICA-Catalonia: Director/Board of the Institute of Health Studies, ID not assigned. Scottish Heart Health Extended Study (SHHEC): Tayside Health Board Dundee District, DM/CL/207.

References

Supplementary materials

  • Supplementary Data

    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.

  • Supplementary Data

    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.

Footnotes

  • Correction notice This article has been corrected since it first published. The open access licence type has been changed to CC BY.

  • Contributors Authors’ contributions: GV, FK, MMF and BH conceived the research question. BH, FK, GV drafted the manuscript along with MMF and HF. GV conducted the statistical analyses. KK, TZ and SB are guarantor of the MORGAM/BiomarCaRE database. HT-P, SSa, VS, BT, FK, MMF, BT, SSo, GC, LI, LP are the principal investigators of the cohorts included in the current analyses. MW, FD, KK, HT-P, SSa, VS, BT, AP, TB, AP, SSo, GC, LI, LP, MB, TZ and SB actively contributed to the interpretation of the results and made critical revision of the manuscript.

  • Funding The BiomarCaRE Project is funded by the EU Seventh Framework Programme (FP7/2007– 2013) under grant agreement HEALTH-F2–2011–278913. The activities of the MORGAM Data Centre have also been sustained by recent funding from EU FP7 project CHANCES (HEALTH-F3–2010–242244). A part of the biomarker determinations in the population cohorts was funded by the Medical Research Council London (G0601463, identification No.80983: Biomarkers in the MORGAM Populations). The MONICA/KORA Augsburg study was initiated and financed by the Helmholtz Zentrum München—German Research Center for Environmental Health, which is funded by the German Federal Ministry of Education and Research (BMBF) and by the State of Bavaria.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.