An inverse association of individual1^–3 and contextual/neighbourhood4 5 socioeconomic status (SES) in adulthood with cardiovascular disease (CVD) has been demonstrated. Low adult SES is associated with poor health behaviours and unfavourable CVD risk factor profiles.3 5^–7 Recent research has focused on the influence of life-course SES on CVD8 9: various study designs have been utilised to examine the impact of SES throughout the life course on adult cardiovascular health and several interrelated theories have been posited.10^–13 Results support a cumulative model of life-course risk, wherein the accumulation of adverse psychosocial and physiological exposures across the life course is associated with increased CVD.9 14^–18

Inquiries into possible mechanisms underlying such associations suggest a sustained, elevated systemic inflammatory burden may mediate social differences in CVD. Chronic and systemic inflammatory upregulation plays a central role in atherosclerosis and its sequelae.19^–23 Lower adult SES is associated with elevated levels of inflammatory markers such as fibrinogen, white blood cell count (WBC), C-reactive protein (CRP) and von Willebrand factor (vWF).24^–26 These and other inflammatory markers and cytokines have in turn been linked with CVD.27^–34 Atherosclerosis begins earlier in life than previously suspected.35 SES plausibly affects health behaviours, nutrition, access to health care, exposure to infection and stress throughout life, thereby influencing an individual’s systemic inflammatory burden and CVD risk.3 7

Early-life SES has been linked with elevated inflammatory marker levels in other studies36^–38 and in this cohort.39 However, no studies have examined the cumulative influence of exposure to adverse SES across life on adult inflammatory levels. In order to explore cumulative adverse SES exposure and systemic inflammation as a potential pathway by which SES and CVD may be linked, we examined the impact of cumulative life-course and adult SES (measured at both the individual and community levels) on adult levels of fibrinogen, WBC, CRP and vWF and overall inflammatory burden in the Atherosclerosis Risk in Communities (ARIC) study.

METHODS

Data

ARIC is a prospective study of atherosclerosis in 15 792 men and women aged 45–64 at baseline (1987–9), sampled from four US communities with an overall 60% participation rate.40 The Life Course Socioeconomic Status, Social Context and Cardiovascular Disease (LC-SES) Study is an ancillary study that retrospectively collected socioeconomic information from childhood (1930–50) and early adulthood (1960–80) on 12 681 African-American and white ARIC participants in 2001–2.41 Individual-level SES and places of residence during childhood and at ages 30, 40 and 50 were collected in ARIC and from annual follow-up telephone interviews.

SES measures

SES was evaluated in childhood (10 years of age), early adulthood (30–40 years of age) and mature adulthood (45+ years). Individual-level SES was evaluated via social class and education; contextual/neighbourhood-level SES was evaluated using SES indices, described below.

Social class was evaluated utilising a derivation of Wright’s schema for categorising social class, an established sociological schema previously examined in epidemiology.42^–44 Participants were assigned a class at each time point based on education and items evaluating ownership, authority and control facets of their job (table 1). Father’s data were used for childhood; principal occupation at 30–40 years of age and current/last occupation were used for young and mature adulthood, respectively. Homemakers were excluded from social class analyses.

Education was evaluated in childhood and adulthood and categorised low/middle/high (table 2). In childhood, father’s education was used; in adulthood, participant’s education was used.

To evaluate neighbourhood SES, participants’ places of residence during childhood and adulthood were linked to US census data (county data for childhood and census tract data for adulthood).45 Of 12 681 participants, 12 373 had US residences in childhood. Of these 12 373, 98.5% (12 187) had their childhood residences successfully linked with US county data. Linkage rates in young adulthood were lower: 66% for 1960 census data, 76% for 1970 data and 85% for 1980 data. A neighbourhood SES index was created for each period (table 3) to evaluate contextual/neighbourhood SES.46

Cumulative social class, education and contextual/neighbourhood SES measures were created by summing the number of times across the life course one was exposed to the lowest category of SES (table 4).14 47

Physiological covariates and outcomes

CVD risk factors, fibrinogen, vWF and WBC were measured at baseline.40 Diabetic status was determined by fasting plasma glucose ⩾126 mg/dl, non-fasting plasma glucose ⩾200 mg/dl, self-reported history of physician-diagnosed diabetes or current medication for diabetes.5 40

Haemostasis variables were analysed in citrated samples,48 fibrinogen by the thrombin time titration method and vWF antigen by enzyme-linked immunosorbent assay.49 Reliability coefficients were 0.72 for fibrinogen and 0.68 for vWF.50 White cell counts were determined using automated cell counters (reliability coefficients 0.96–1.00).

CRP measures were obtained from fasting blood taken from the 5552 participants in the Dental ARIC study during the 1996–8 ARIC Visit (eligibility required having teeth, no medical contraindications to periodontal probing and consent).51 CRP was quantified using ELISA assay (range of detection 0.5–50 mg/l).52 Intra-plate and inter-plate coefficients of variation ranged from 0–37.3% and 0–10.6%, respectively.

Statistical methods

Incomplete earlier-life address, education and census data resulted in high missing data rates for cumulative education (16.4%), class (28.1%) and neighbourhood SES (36.6%) measures for the overall cohort. Multiple imputation (MI) was used to reduce bias.53 Data were considered to be missing at random (MAR), an assumption that usually holds in self-reported questionnaire data and when erroneous data tend to have only a minor impact on estimates.54 55 Multivariate imputation by chained equations (MICE) was carried out in Stata 8.2 SE utilising 29 demographic covariates and five imputations of 10 iterations.56

Inflammatory outcomes were log-transformed. Overall inflammatory burden was assessed using an inflammation score ranging from 0 to 4 (one point was assigned for each of the four inflammatory markers with a level above its race/ethnicity-specific median). This score has been previously utilised and demonstrated a graded, statistically significant association with risk of incident diabetes.57 Analyses using CRP or the inflammatory score as outcomes were limited to the 5022 dental ARIC study participants with early-life SES data.

We initially fitted multi-level models with random intercepts for each tract or county to account for residual within-area correlation. However, as they did not significantly improve model fit, we subsequently used linear regression to examine the impact of each SES measure on each inflammatory outcome.

One set of race/ethnicity-specific models (model A) adjusted for age, gender and ARIC centre. A second set (model B) additionally adjusted for high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), body mass index (BMI), hypertension (HTN) and HTN medication, family history of CVD, diabetes status, alcohol intake, smoking status and cigarette-years smoked. For cumulative SES measures, the referent was zero exposures to adverse life-course SES conditions; for adult SES measures, the referent was highest SES.

Geometric mean levels of inflammatory markers were computed for each level of SES by exponentiating parameter estimates. Statistical interaction between SES and potential confounders was assessed using partial F tests.58 Tests for linear trend were run using interval SES variables.

RESULTS

African-American participants had lower scores for adult and cumulative SES measures, higher mean fibrinogen, CRP and vWF levels, and lower mean WBC levels (all p<0.001) than white participants (table 5). Patterns were similar across gender groups. Percentages and standard errors are presented for SES measures as these are composite estimates from five multiply imputed datasets. Neighbourhood SES categories were based on race/ethnicity-specific distributions and cannot be directly compared across race/ethnicity groups.

Cumulative life-course SES, white participants

Among white participants, increasing cumulative life-course exposure to lower SES conditions for all SES measures was associated with elevated levels of each adult inflammatory marker in minimally adjusted models (table 6; vWF results in supplementary table 1). Graded increments in fibrinogen and WBC levels were observed with greater adverse SES exposure (p for trend <0.01 for individual-level SES measures in minimally adjusted models).

Greater cumulative exposure to unfavourable SES was strongly associated with elevated fibrinogen and WBC (table 6). For example, white participants with three versus zero life-course exposures to the worker social class group had geometric mean fibrinogen of 296.3 mg/dl (95% CI: 292.4 to 300.2), versus 283.9 mg/dl (95% CI: 281.8 to 286.1) (p<0.001).

Lower categories of all measures were associated with graded elevations in the overall inflammatory score (supplementary table 2). Differences between lowest and highest individual-level SES categories all attained statistical significance in minimally adjusted models, suggestive of a greater overall inflammatory burden (p for trend <0.01).

While CVD risk factor adjustment attenuated all cumulative SES-inflammation associations, increasing adverse life-course exposures remained associated with higher CRP, fibrinogen and vWF (table 6). For fibrinogen, the associations for the most-exposed versus least-exposed groups remained statistically significant in all cumulative SES measures.

Adult SES, white participants

Lower adult SES was strongly associated with elevated inflammatory levels among white participants (table 6; vWF results in supplementary table 1). Graded elevations in inflammatory markers were observed with decreasing SES for all adult measures (p for trend <0.001 for fibrinogen and WBC in minimally adjusted models). For every adult SES measure, participants in lowest versus highest SES groups had statistically significantly higher levels of all markers; the sole exception was for Worker class adults and CRP. As with cumulative SES, fibrinogen and WBC were strongly associated with adult SES.

Lower categories of all measures were associated with graded elevations in the overall inflammatory score (supplementary table 2); differences between lowest and highest SES categories were all significant in minimally adjusted models. Relative increments in inflammatory score by SES were similar in size to those observed within levels of cumulative SES.

CVD risk factor adjustment attenuated most associations (table 6). However, low education remained significantly associated with elevated fibrinogen and WBC. Elevated fibrinogen also remained significantly associated with lowest social class and neighbourhood SES. Associations with the inflammatory score were no longer statistically significant after covariate adjustment, but graded associations remained.

Cumulative SES, African-American participants

Among African-Americans, increasing cumulative exposure to adverse SES was inconsistently associated with elevated inflammatory levels (table 7; vWF results in supplementary table 1). Greater life-course exposure to lower education and social class was associated with elevated fibrinogen and vWF (and CRP for social class) in minimally adjusted models; greater exposure to low-SES neighbourhoods was associated with elevated CRP and fibrinogen. Associations were weaker and less consistent among African-American participants than among white participants. Stepwise, graded elevations in inflammatory marker levels were not consistently observed by cumulative SES category, as they were in white participants. Only slight increments in the overall inflammatory score by greater exposure to adverse SES conditions were observed (supplementary table 2).

CVD risk factor adjustment strongly attenuated the cumulative SES-inflammatory marker associations. Moreover, certain associations were inverted after CVD risk factor adjustment, with greater exposure being associated with lower inflammatory levels.

Adult SES, African-American participants

Lower adult SES was associated with elevated inflammatory levels among African-Americans (table 7; vWF results in supplementary table 1). Patterns were more consistent than for cumulative SES. Levels of all markers were higher for lowest versus highest category of all SES measures; graded associations were observed for fibrinogen and WBC.

The influence of lower adult SES on inflammation was generally smaller among African-American participants than white participants. For example, the relative increments in CRP, fibrinogen, WBC and vWF observed for lowest versus highest education among African-American participants were 18%, 4%, 0% and 7%, respectively, versus 37%, 6%, 10% and 5% for white participants, respectively. Low adult SES was most notably associated with adult fibrinogen, where graded associations were observed for all adult SES measures (p for trend <0.05).

Lower adult SES was modestly associated with an elevated overall inflammatory score among African-Americans (supplementary table 2). Associations were smaller than for white participants for individual-level measures, but larger for neighbourhood-level SES (p for trend <0.05).

As for cumulative SES, CVD risk factor adjustment substantially attenuated (and in some instances inverted) most associations (table 7).

DISCUSSION

Cumulative life-course and adulthood SES were markedly, inversely associated with adult inflammatory marker levels among white people. Fibrinogen and WBC were most strongly associated with life-course and adult SES; statistically significant, linear, graded increments in these markers’ levels were observed with decreasing SES. CRP and vWF were also inversely associated with SES, although associations did not consistently attain statistical significance. Results among African-Americans were less clear: patterns of association were inconsistent, particularly in covariate-adjusted models. Adult SES was generally more strongly associated with inflammatory levels than life-course SES. Graded, stepwise increments in the overall inflammatory score observed with worse SES (both cumulative life-course and adult measures) underline the consistency of the analytical results, particularly among white people.

Weaker, less consistent associations among African-Americans may be due to smaller sample size and power. Also, race/ethnicity-specific SES distributions differed significantly; the range of SES levels among African-American participants was (1) composed of lower SES scores and (2) narrower than the range among white participants. Potentially relevant aspects of SES (for example, healthcare access, experience with racism) were not available. Associations may be patterned differently by race/ethnicity, influenced by differential health selection by race/ethnicity or may point to limitations present in the evaluation of SES among African-Americans in ARIC.6 46 59

Studies have consistently reported associations between the number of adverse life-course SES exposures and levels of CVD risk factors and events.9 Our results support these findings and further suggest that earlier-life exposures to adverse conditions are linked to elevated adult inflammation.

A number of mechanisms may plausibly underlie this link. The fetal origins hypothesis suggests that malnutrition and other physiological insults during critical developmental periods may lead to adult metabolic dysfunction associated with pro-inflammatory states.60 Childhood exposure to adverse SES conditions during sensitive developmental periods may have similar effects. Low SES has been associated with adult health behaviours linked to inflammation,2 7 and life-course SES may be a strong determinant of adult health behaviours given the early development of behaviour patterns. Stress, depression and/or hostility linked to low SES may lead to chronic activation of the hypothalamic-pituitary-adrenocortical axis, increased release of corticosteroids, abnormal platelet function and other conditions.3 Finally, low SES conditions/environments plausibly increase exposure to potentially pro-inflammatory or infectious agents such as microbial pathogens, ambient air pollutants and adverse work conditions.61

Childhood contextual SES was evaluated at the county level while adult contextual SES was evaluated at the census tract level; the impact of early-life contextual SES is therefore probably underestimated. Additionally, the cumulative life-course model utilised assumes that adverse SES exposures have the same impact regardless of length or timing of exposure and does not address the potentially differential impact of upward or downward social mobility.9 Selective survival of healthier participants, a 60% baseline participation rate and the availability of CRP only among healthier participants limited the socioeconomic and health diversity of participants and may have biased results towards the null. Errors in remembering early-life SES may also have biased results towards the null.62

Strengths of this study include use of SES measures from three life epochs, a separate analysis of adult SES, examination of both individual-level and contextual-level SES and the availability of multiple inflammatory markers. Additionally, the use of MI represents a systematic effort to minimise biases due to missing data.

While the stronger association between adult SES and inflammation is unsurprising, inflammatory levels were closely tied to life-course SES. The graded relations between the number of life-course exposures to low SES and inflammatory levels suggest that the impact of these exposures is cumulative. These findings point to a sustained increase in inflammation as a response to adverse life-course SES, matching the development of chronic vascular disease in young adulthood and progressing throughout life.35 Our results suggest a framework linking adverse life-course SES exposures to end-organ pathology in Western industrialised countries. These findings have implications for our understanding of the long-term development and progression of chronic vascular disease and speak to a need for further examination of the earlier-life development of CVD risk factors and poor health behaviours.