Abstract
Background
Childhood socioeconomic status (SES) is related to risk for cardiovascular disease in adulthood, perhaps in part due to associations with inflammatory and hemostasis processes. We tested the hypotheses that childhood SES is related to C-reactive protein (CRP), fibrinogen, Factor VIIc, and plasminogen activator inhibitor-1 (PAI-1) in mid-life women and that the associations are mediated by adult SES and/or adult body mass index (BMI).
Methods
Using data from the prospective Study of Women’s Health Across the Nation, we classified 1067 black and white women into three multidimensional childhood SES groups based on latent class analysis. Biological measures were assessed across 7 years along with covariates and mediators, and analyzed by mixed regression models, followed by tests for mediation.
Results
Compared to women raised in high SES families, those from the lowest SES families had higher levels of CRP, b (standard error) = .37 (.11), PAI-1, b = .23 (.07), Factor VIIc, b = .05 (.02), and fibrinogen, b = 11.06 (4.89), after adjustment for ethnicity, site, age, ratings of health between ages 11 and 18, visit, smoking status, menopausal status, stroke or heart attack, medications, and hormone use. Introduction of adult SES and BMI into the models reduced the childhood SES associations to nonsignificance for all four measures. Indirect mediation was apparent for adult education and BMI for CRP, and BMI for PAI-1.
Conclusions
Women raised in lower SES families had elevated markers of inflammation and hemostasis, in part due to elevated BMI and education in adulthood.
Keywords: socioeconomic status, childhood, inflammation, hemostasis, race, women
BACKGROUND
Altered inflammation and hemostasis, two inter-related systems, play major roles in the pathophysiology of atherosclerosis and other vascular diseases. Hemostasis is the arrest of bleeding in response to injury. An abnormal response can lead to thrombosis or excessive bleeding, depending on the balance of multiple systems. Epidemiological studies have demonstrated the associations between biological measures of inflammation and hemostasis with risk for cardiovascular disease (CVD), with the evidence most substantial for C-reactive protein (CRP) and fibrinogen (1, 2). CRP is a protein synthesized in the liver as an acute phase reactant and is a nonspecific marker for inflammation, infection, and tissue damage (3). Fibrinogen is a glycoprotein synthesized in hepatocytes and is part of the platelet aggregation and coagulation process (4). There is less evidence of relations between other markers of inflammation/hemostasis and CVD outcomes. Plasminogen-activator inhibitor type I (PAI-1) is synthesized in platelets and endothelial cells and serves as a major inhibitor of fibrinolysis, which should lead to an increase in arterial fibrin deposition and thrombosis (5). Factor VIIc is part of the coagulation cascade but its role as a risk factor for CVD is controversial (6).
Accumulating data suggest that family SES in childhood is related to risk for CVD (7), and understanding the mechanisms underlying these relations has been a topic of intense interest. A prominent hypothesis is that low childhood SES increases risk for inflammation and hemostasis, which, in turn, increases CVD risk. Large scale studies have investigated the association of childhood SES and inflammation and hemostasis in adulthood. However, the results vary by measure of inflammation, measure of SES, demographic group, and covariates. Lower status parental occupation was related to higher CRP among adults in their thirties (8) and among the elderly (9). In the Atherosclerosis Risk in Communities Study, lower status parental occupation and education were related to higher fibrinogen levels and white blood cell count, particularly among whites (10). In other work, lower parental education was related to higher CRP levels among women but not among men (11). Brazilian men and women whose mothers had low education had elevated CRP (12). In a separate Brazilian sample, men whose families had higher income had elevated CRP and women whose mothers had low education had elevated CRP (13).
Investigators have questioned whether childhood SES is related to adult inflammation and hemostasis at least in part because low childhood SES leads to less education in adulthood, which leads to poor health behaviors in adulthood. Low childhood SES may be related to the development of obesity, which, in turn, may lead to a less favorable inflammatory and hemostasis profile (9). Few studies find that childhood SES has its effects on inflammation/hemostasis, independent of adult SES (e.g., (11)) but others report nonsignificant results, after adjusting for adult SES (12–14). Some studies consider body mass index (BMI) as one of a group of cardiovascular risk factors so it is difficult to infer the specific mediational role of BMI in these models (10, 14, 15). One study controlled for birth weight and childhood BMI but not adult BMI, with the association remaining between childhood SES and adult CRP levels above 3 mg/dl (8).
Based on data from the Study of Women’s Health across the Nation (SWAN), the present report examines whether childhood SES is associated with four measures of inflammation and hemostasis: CRP, fibrinogen, Factor VIIc, and PAI-1. Rather than examine individual aspects of family SES in childhood, women were classified into groups identified by latent class analysis using parental education and family assets. The latent class approach has advantages because it improves reliability and combines indices of status and resources, the major conceptual dimensions of SES, into one index. In addition to offering a combined childhood index of SES, the present study adds to the literature in a number of ways. This study includes a comprehensive set of inflammatory and hemostasis measures and examines the associations prospectively across a 7-year period when the biological measures were repeatedly assessed. It explores whether the associations vary by race or ethnicity. Finally, it tests the influence of several mediators: adult SES and BMI in adulthood. It is important to study women in particular because they have higher levels of inflammation than do men in midlife, there are unique determinants of inflammation in women, e.g. use of hormone therapy (16–18), and the need to enhance understanding of cardiovascular risk in women in the peri- and post-menopausal years.
METHODS
Sample
The Study of Women’s Health Across the Nation (SWAN) is a multi-site, community-based, prospective study of aging and the menopause transition. From a cross-sectional study of 16,065 women from seven locations (Boston, Chicago, the Detroit area, Los Angeles, New Jersey, Oakland, and Pittsburgh), 3302 women who at the 1996 baseline were aged 42–52 years, had menstruated in the past 3 months, had an intact uterus and at least one ovary, were not using oral contraceptives or hormone therapy, and self-identified with designated race-ethnic groups were recruited for the prospective study. Details on sampling and recruitment methods are published elsewhere (19). Annual study visits between baseline and follow-up year 13 included interviews, weight, height, and waist circumference, questionnaires, and a blood draw. The analytical sample for the present study includes participants from the four SWAN locations (Boston, Chicago, Detroit, and Pittsburgh) that administered the Childhood Context Ancillary Study questionnaire during visit 13. These sites enrolled black and white women. Among the 1399 SWAN participants across these sites, 1109 returned the ancillary questionnaire (response rate = 79%). From these participants, 30 were excluded due to having had a history of stroke or heart attack at baseline. Among the remaining women we had inflammation and hemostasis data and relevant covariates for 1067 women. SWAN protocols were approved by the institutional review boards at each site, and each participant provided written informed consent.
Inflammation and Hemostasis Measures
Phlebotomy was performed in the morning after an overnight fast. Blood was maintained up to 1 hour at 4 C until separated and then frozen and sent on dry ice to the Medical Research Laboratory (Lexington, KY) for analysis. For budgetary reasons, assays were completed at baseline and at SWAN study years 1 and 3–7 for hsCRP, and PAI-1, and at baseline and SWAN years 1, 3, 5, and 7 for Factor VIIc and fibrinogen. All assays were measured in plasma. The hsCRP was quantitated using an ultrasensitive rate immunonephelemetric method (Dade-Berhing, Marburg Germain). The sensitivity of the assay was .03 mg/dL and the coefficients of variation (CVs) at the concentrations of .05 and 2.2 mg/dL were 10% to 12%, and 5% to 7%, respectively. Fibrinogen and Factor VIIc were measured in frozen citrated plasma (MLA ELECTRA 1400C; Medical Laboratory Automation Inc. Mt. Vernon, NY) using a turbidometric detection system. Fibrinogen monthly interassay CVs were 2.3% to 3.5% and 2.6% to 3.6% at mean concentrations of 250 and 140 mg/dL, respectively, and Factor VIIc monthly interassay CVs were 7.8%, 5%, and 4% for mean activities of 8%, 45%, and 99% respectively. PAI-1 was measured with a sandwich procedure using a solid-phased monoclonal antibody and enzyme-labeled goat second antiserum for detection (IMUBIND plasma PAI-1 enzyme-linked immunosorbent assay (ELISA); American Diagnostica, Greenwich CT). PAI-1 monthly interassay CVs were 5% to 9% and 4% to 9% at mean concentrations of 7 and 22.5 ng/dL, respectively. Consistent with published guidelines, hs-CRP values greater than 10 were excluded (11.8% of observations). All values except for fibrinogen were skewed and subjected to log transformation.
Childhood Circumstances
Information on childhood circumstances was collected with the 10-item Childhood Context Ancillary Study questionnaire that was administered during visit 13 in 2012/2013. To assess childhood SES, it asked participants about their mother’s and father’s highest educational level in 6 categories (e.g. less than high school, post-graduate degree). They were asked whether their family owned a car (yes/no) or home (yes/no) when they were children or teenagers. They were also asked whether their childhood family ever received public assistance (yes/no) and ever had difficulty paying for food or rent or making ends meet (yes/no) separately when they were less than 11 or 11 to 18 years. These were combined into yes in either time period versus no in all periods to be comparable in time frame to the other questions. In addition to the socioeconomic information, the questionnaire asked participants about their overall health when they were less than 11 and between 11 and 18 in 5 categories (excellent, very good, good, fair and poor). Participants were grouped into women reporting their childhood health as fair or poor health, very good or good, or excellent health at age 11 – 18. All questions with the exception of perceived child health questions had been administered at SWAN year 07 to 259 women at the Pittsburgh site. The concordance of the individual questions from year 07 to year 13 varied from 86.1 to 96.9% among these participants.
Adult Mediators and Covariates
We included race/ethnicity (black vs white), site, baseline age, and perceived health at ages 11 – 18 as time-invariant covariates. Time varying covariates during the follow-up period were visit, current smoker (yes/no), menopausal status (postmenopausal, i.e., natural or surgical, late perimenopause, early perimenopause, premenopause, unknown, i.e., missing data or use of hormone therapy prevented accurate assessment of final menstrual period), heart attack or stroke, use of medications affecting inflammation and hemostasis (yes/no), and use of hormone therapy (yes/no). As potential mediators of the effects of childhood SES class, we included two indicators of adult SES measured at baseline, adult educational status (high school degree or less, some college, 4 year college degree or more) and financial hardship (very, somewhat, or not hard to pay for basics), and BMI (time varying).
Statistical analysis
We identified distinctive subgroups of SWAN participants based on their childhood SES information using latent class analyses following recommended procedures in Mplus (20, 21). We used mixed model linear regression using childhood SES class groups and covariates (time invariant: ethnicity, site, baseline age, ratings of health between ages of 11 and 18, and time varying: visit, smoking status, menopausal status, stroke or heart attack, medications, and hormone use) to predict each outcome separately across the 7 years of follow-up (Model 1). The predictor variable was childhood SES group, with highest SES group as the referent. P-values < .05 for childhood SES group were considered statistically significant.
Using the same covariates in Model 1, analyses evaluated the association between childhood SES group and BMI at the same visits of the outcome measures, i.e., 0, 1, and 3 −7. Then we conducted two additional models: one including two indicators of SES, adult education and adult financial hardship (Model 2), and another adding adult education, financial hardship, and BMI (Model 3).
Mediational analyses varied for adult education and BMI. Since adult education was a categorical variable, ordinal logistic regression was used to test the relation between childhood SES latent class and adult education, as the dependent variable. Then the association between SES and outcome (e.g., logged CRP) while adjusting for adult education and other covariates was assessed by linear regression. Since the coefficients from ordinal logistic regression and linear regression had different scales, the coefficients were standardized (i.e., coefficient/standard error) so that they could be used in the analysis. Then the transformed coefficients were multiplied together to obtain the indirect effect (24). The indirect effects for BMI were obtained using linear regression as BMI was a continuous variables, with the effects of these variables estimated by the product of coefficients method (25). Confidence intervals of the indirect effects were estimated by bootstrap method with 1000 replications (26). We also tested for interactions between childhood SES classes and adult education groups but none was statistically significant. All models were estimated with SAS version 9.3.
RESULTS
Sample Characteristics
Approximately 56% of the sample was white and 44% black (Table 1, left column), and 21% had a high school degree or less and 44% had a four year college degree or more (Table 2, left column). The number of smokers was small and on average the sample was overweight to obese.
Table 1.
Distribution of demographics and childhood socioeconomic indicators among overall sample and within three latent classes of childhood socioeconomic conditions
| Full Sample | Distribution within Latent Class
|
|||
|---|---|---|---|---|
| 1 (“adverse”) Parents low-educated and poor |
2 (“fair”) Parents low-educated, not poor |
3 (“good”) Parents high-educated, not poor |
||
| N | 1067 | 271 | 539 | 257 |
| Latent class membership (% columns sum to 100) | ------ | 25.4 | 50.5 | 24.1 |
| Father’s education (%) | ||||
| Less than high school | 28.3 | 35.1 | 37.1 | 2.7 |
| High school credential | 28.4 | 22.9 | 43.0 | 3.5 |
| Technical or vocational school | 5.1 | 1.5 | 9.1 | 0.4 |
| Some college | 10.5 | 7.4 | 9.3 | 16.3 |
| Bachelor’s degree | 11.1 | 4.1 | 0.0 | 41.6 |
| Postgraduate degree | 8.7 | 2.2 | 0.0 | 33.9 |
| Don’t know | 8.0 | 26.9 | 1.5 | 1.6 |
| Mother’s education (%) | ||||
| Less than high school | 26.5 | 40.6 | 30.8 | 2.7 |
| High school credential | 39.0 | 33.6 | 51.2 | 19.1 |
| Technical or vocational school | 7.1 | 4.4 | 8.4 | 7.4 |
| Some college | 11.3 | 5.5 | 8.7 | 22.6 |
| Bachelor’s degree | 7.7 | 2.2 | 0.6 | 28.4 |
| Postgraduate degree | 4.8 | 1.1 | 0.0 | 18.7 |
| Don’t know | 3.7 | 12.6 | 0.4 | 1.2 |
| Childhood family (%) | ||||
| …did not own a car | 13.1 | 39.9 | 4.9 | 2.0 |
| …did not own a home | 20.1 | 63.1 | 6.2 | 4.3 |
| …ever received public assistance | 18.3 | 62.1 | 4.8 | 0.8 |
| …ever had difficulty paying for food or rent | 35.4 | 82.8 | 21.4 | 15.2 |
| Race (%) | ||||
| White | 55.8 | 31.7 | 57.7 | 77.0 |
| Black | 44.2 | 68.3 | 42.3 | 23.0 |
| SWAN location (%) | ||||
| Boston | 26.1 | 25.8 | 21.0 | 37.0 |
| Chicago | 14.3 | 13.7 | 13.5 | 16.3 |
| Detroit | 31.8 | 39.1 | 35.3 | 16.7 |
| Pittsburgh | 27.9 | 21.4 | 30.2 | 30.0 |
Table 2.
Inflammation and hemostasis characteristics and covariates at baseline in the full sample and according to three latent classes of childhood socioeconomic status. (N=1067)
| Classes
|
|||||
|---|---|---|---|---|---|
| Total Sample | Low parental education, poor | Low parental education, not poor | High parental education, not poor | P-value | |
| Baseline | |||||
| CRP mg/l, Median (IQR) | 1.6 (0.6, 3.9) | 2.1 (0.8, 4.6) | 1.7 (0.6, 4.0) | 1.1 (0.5, 2.9) | <.0001 |
| Fibrinogen mg/dl, Mean (SD) | 301.5 (67.3) | 312.4 (66.3) | 302.5 (67.7) | 287.6 (66.5) | <.0001 |
| Factor VIIc %, Median (IQR) | 115.0 (98.0, 132.0) | 117.0 (98.0, 136.0) | 114.0 (98.0, 132.0) | 113.0 (96.0, 128.0) | .60 |
| PAI-1 ng/ml, Median (IQR) | 21.1 (12.3, 35.4) | 23.4 (15.6, 37.2) | 19.7 (12.0, 34.5) | 20.4 (10.4, 33.5) | .001 |
| Smoker, N (%) yes | 176 (16.7) | 66 (24.6) | 83 (15.6) | 27 (10.6) | <.0001 |
| Body Mass Index, Mean (SD) | 29.8 (7.7) | 32.0 (8.6) | 29.7 (7.3) | 27.9 (6.8) | <.0001 |
| Menopausal Status, N (%) | .003 | ||||
| Premenopausal | 570 (54.3) | 123 (45.7) | 298 (56.1) | 149 (59.8) | |
| Early Perimenopausal | 479 (45.7) | 146 (54.3) | 233 (43.9) | 100 (40.2) | |
| Adult Education, N (%) | |||||
| High school or less | 219 (20.8) | 83 (31.0) | 121 (22.7) | 15 (5.9) | <.0001 |
| Some college | 371 (35.2) | 109 (40.7) | 204 (38.4) | 58 (22.8) | |
| 4 year college degree and higher | 464 (44.0) | 76 (28.4) | 207 (38.9) | 181 (71.3) | |
| Financial hardship, N (%) | <.0001 | ||||
| Hard | 76 (7.1) | 35 (12.9) | 28 (5.2) | 13 (5.1) | |
| Somewhat hard | 293 (27.5) | 93 (34.3) | 154 (28.6) | 46 (17.9) | |
| Not at all hard | 698 (65.4) | 143 (52.8) | 357 (66.2) | 198 (77.0) | |
P-values are unadjusted and are based on ANOVA or chi square.
Latent class analyses revealed that the women could be classified into 3 childhood SES groups. Class 1 contained 25 of the women and is the lowest SES, consisting of women with low-educated and poor parents, e.g., family on public assistance (Table 1). Class 2 contained 51% of women with low educated but nonpoor parents. Class 3 contained 24% of the women and is the highest SES, with high educated and nonpoor parents. The analyses identified the 3-class model as having the best fit. It had the lowest Bayesian Information Criterion score (BIC1-class=12112, BIC2-class =11545, BIC 3-class=11453, BIC4-class =11473). In addition, the Vuong-Lo-Mendell-Rubin Likelihood Ratio Test indicated that 2-classes are significantly better than 1-class (p<0.001), 3-classes are better than 2-classes (p=0.076), and 4-classes make no further improvement (p=0.816). The 3-class model also provides good classification quality. It has an entropy score of 0.712 [0.80 is considered high, 0.60 medium, and 0.40 low (22)]. In addition, the correct class assignment probabilities are 0.86 for the lowest SES group, 0.85 for the middle SES group and 0.93 for the highest SES group, which exceed the 0.70 recommended threshold (23). Because the entropy score was less than .80, we focused below on effects that were significant at p < .01, but report findings that are significant at p < .05.
The three childhood SES groups varied at baseline on smoking status, BMI, and initial menopausal status (Table 2). Women in the lower childhood SES group were more likely to be smokers and have higher BMI. Women in the lowest and highest childhood SES groups tended to be early perimenopausal compared to those in the middle group.
Women varied by childhood SES class in their levels of CRP, fibrinogen, and PAI-1 at baseline (Table 2). The values for CRP and fibrinogen increased monotonically with decreasing childhood SES, P ≤ .0001 from linear trend analyses. Baseline Factor VIIc was not related to childhood SES groups. The covariation of the individual indices that comprise the childhood latent classes with the measures of inflammation, hemostasis, and adult covariates are also presented in Table S1, Supplemental Digital Content 1.
All hemostasis and inflammatory measures were correlated. At baseline CRP was associated with Factor VIIc, r = .26, fibrinogen, r = .59, and PAI-1, r= .40; Factor VIIc was associated with fibrinogen, r = .12, and PAI-1, r = .29, and fibrinogen was associated with PAI-1, r = .19, all P ≤.0001.
Associations between Childhood SES Class Groups and Inflammation, Hemostasis, and Adult Mediators
Independent of covariates, lower childhood SES was associated with higher CRP and PAI-1; the trends for fibrinogen and Factor VIIc approached significance (Table 3, Models 1). The estimates for the low parental education/poor group differed significantly from high parental education/nonpoor group, but neither differed from the intermediate group (data not shown). Interactions with race and childhood SES class groups showed no significant relationships. Interactions of childhood SES class groups with visit number (0, 1, 3 – 7) showed no significant relationships for CRP or PAI-1, suggesting the change over time in CRP or PAI-1 did not vary by childhood SES class groups.
Table 3.
Estimates from multiple regression mixed models of childhood socioeconomic status predicting inflammation and hemostasis (N=1067)
| Model 1 | Model 2 | Model 3 | ||||
|---|---|---|---|---|---|---|
|
| ||||||
| b (SE) | p-value | b (SE) | p-value | b (SE) | p-value | |
| C-Reactive Protein | .001 | .03 | .31 | |||
| Low parental education, not poor | .27 (.09) | .002 | .19 (.09) | .04 | .11 (.07) | .13 |
| Low parental education, poor | .37 (.11) | .001 | .27 (.11) | .01 | .09 (.09) | .34 |
| Fibrinogen | .08 | .23 | .84 | |||
| Low parental education, not poor | 5.24 (4.08) | .20 | 3.38 (4.19) | .42 | 1.22 (3.77) | .75 |
| Low parental education, poor | 11.06 (4.89) | .03 | 8.54 (5.07) | .09 | 2.66 (4.58) | .56 |
| Factor VIIc | .07 | .27 | .60 | |||
| Low parental education, not poor | .02 (.02) | .15 | .01 (.02) | .44 | .007 (.02) | .68 |
| Low parental education, poor | .05 (.02) | .02 | .03 (.02) | .11 | .02 (.02) | .33 |
| PAI-1 | .002 | .02 | .18 | |||
| Low parental education, not poor | .14 (.06) | .01 | .12 (.06) | .04 | .08 (.05) | .11 |
| Low parental education, poor | .23 (.07) | .001 | .20 (.07) | .005 | .11 (.06) | .09 |
Model 1. Ethnicity, site, ratings of health at ages 11–18 years, baseline age, visit, smoking, menopausal status, heart attack or stroke, medications affecting outcomes, hormone use.
Model 2. Model 1 + adult education and adult financial hardship
Model 3. Model 2 + BMI
Note: p-values listed next to outcome reflect overall differences among the 3 class groups; p-values listed next to class group compare the specific group to the referent group, high parental education and not poor.
Childhood SES groups varied in their education attainment reported at baseline (Table 2). Those in the low parental education/poor group were least likely to attain a college degree, whereas those in the high parental education/non poor group were most likely to attain a four year degree or higher. They also varied in reports of how hard it is to pay for basics, with those from families with low parental education/poor being most likely to report that it was hard.
Independent of covariates included in Model 1, neither adult education groups nor adult financial hardship was related to fibrinogen, p = .77 and p = .17, or to PAI-1, p = .25, and p = .48, respectively. Adult education groups were related to CRP, p = .04, and Factor VIIc, p = .04. Adult financial hardship was unrelated to CRP in the same model with education, p = .10, and tended to be related to Factor VIIc, p = .06. Compared to women with a college degree, women with a high school degree or less tended to have higher CRP levels, estimate = .19 (SE .10), p = .06, and women with some college had higher CRP levels, estimate = .20 (SE .09), p = .02, and Factor VIIc levels, estimate = .04 (.02), p = .01.
Independent of covariates included in Model 1, childhood SES class groups varied in BMI at years 0, 1, and 3 −7, the visits concurrent with the biological measurement, p = .001. Compared to the highest childhood SES group, women in the lowest childhood SES group had higher adult BMI over time, estimate = 2.13 (SE=.67), p = .002, as did women in the middle childhood SES group, estimate = 1.09 (SE=.56), p = .05.
Influence of Adult Education and Body Mass Index on the Associations between Childhood SES Class Groups and Biological Measurements (Models 2 and 3)
Adjustments for adult education and financial hardship attenuated the relationships of childhood SES, with associations nonsignificant for Factor VIIc and fibrinogen (Table 3, Model 2). Financial hardship was not related significantly to any of the outcomes in Model 2. Further adjustments for BMI reduced the relationship of childhood SES group and CRP and PAI-1 to nonsignificance (Table 3, Model 3). BMI was a highly significant predictor of all measures of inflammation and hemostasis, all ps < .0001 in Model 3 (data not shown).
Mediational tests showed that indirect effects of adult education groups with adult financial hardship and BMI in the model were significant for CRP, 14.74 (95% CI 2.76 – 28.30), and nonsignificant for PAI-1. The indirect effects of adult BMI were significant for CRP, 0.09 (95% CI .02 – .16), and for PAI-1, 0.05 (95% CI .02 – .09), with adult education groups and financial hardship in models. Mediational analyses were repeated with BMI from the immediately prior visit as a covariate on CRP and PAI-1 at visits 1, 3 – 7. The indirect effects for prior BMI were also significant, 0.07 (95% CI .01-.13) for CRP, and 0.04 (.004 – .07) for PAI-1.
Comparison of b’s for childhood SES class showed that for CRP they were reduced by 23% with adult education (treated as a pseudo-continuous variable) and by an additional 69% with BMI. The childhood SES class estimate for PAI-1 was reduced from model 2 to model 3 by 45% with BMI.
DISCUSSION
The present analysis showed that midlife women classified into one of three groups based on parental education and family assets varied in inflammation and hemostasis, with the women whose parents had little education and few assets having the highest levels. Adjustments were made for ethnicity, site, baseline age, perceived health at ages 11 – 18, visit, smoking, menopausal status, reports of heart attack or stroke, medications affecting outcomes, and hormone use. There was no evidence that the relationships varied by racial/ethnic status. Thus, childhood SES group was a strong predictor of measures of inflammation and hemostasis in midlife women.
We addressed statistically whether obtained associations might be due to women from lower childhood SES families achieving less education in adulthood and/or whether these women were more likely to have a higher BMI by adulthood. Our analyses suggested that childhood social class group was strongly related to adult education, which partially mediated the associations with CRP. Further adjustment by adult BMI concurrent with measures of CRP and PAI-1 reduced the effects of childhood SES class to nonsignificance and accounted for a substantial portion of the childhood social class group estimates. Thus, these findings suggest that exposure to early childhood SES leads to both low educational attainment and risk for obesity in adult women, which, in turn, leads to elevated inflammation and hemostasis.
Our study had a number of limitations. Our childhood SES measures were retrospectively reported and collapsed across developmental stages. However, prior studies report that retrospective measures of childhood SES are reliable and accurate (27, 28). In our own sample at one site, assessments of childhood SES five years apart were remarkably consistent. Our findings should not be extrapolated to other ages or to men. Because we did not measure childhood weight and height, we could not detect when prior to mid-life the association of BMI and inflammation and hemostasis would begin to emerge. Note, however, that we did statistically control for the women’s ratings of their health at ages 11–18 years, which is a more general indicator of health, albeit retrospective. Although we included a number of covariates in the model, some potentially significant covariates were not included, e.g. asthma and physical activity.
We also did not have the data to access when in the life span prior to mid-life associations of childhood SES and inflammation and hemostasis would emerge. An enumerative review concluded that the association between childhood SES and CRP in childhood and adolescence was inconsistent, with 2 positive studies out of 7 (29). This suggests that the associations may emerge in young adulthood or later. Earlier associations may also be apparent with more adverse childhood experiences. For example, one of the four SWAN sites had collected information regarding exposure to childhood abuse and neglect and found that independent of adult education, a history of child abuse was related to elevated CRP levels in part through adult obesity (30). Similarly in the Dunedin Study maltreatment in childhood prospectively measured was related to high CRP levels in young adult men and women (31). Several reviews of the literature suggest that exposure to interpersonal violence is related to the development of obesity, with associations being somewhat stronger in women than in men (32, 33). Future research is needed to examine the types, intensity, and timing of adverse childhood experiences in relation to development of abnormal levels of inflammation and hemostasis in adulthood.
Our study has a number of strengths. We used repeated measures that reflect both hemostasis and inflammatory processes. We followed the women over 7 years and had extensive assessments of relevant covariates. Our analyses were able to show indirect effects of obesity and adult education in understanding the relationship of childhood SES and inflammation.
In summary, we found that black and white women who were raised in poor, less educated families had elevated levels of hemostasis and inflammation in adulthood. The associations for childhood SES were in part mediated by adult education and BMI. Prevention of obesity in low SES households may be one important strategy to reduce the negative consequences of low childhood SES on risk for CVD in adult women.
Supplementary Material
Acknowledgments
The Study of Women’s Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), Department of Health and Human Services, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women’s Health (ORWH) (Grants U01NR004061; U01AG012505, U01AG012535, U01AG012531, U01AG012539, U01AG012546, U01AG012553, U01AG012554, U01AG012495). This manuscript was also supported in part by the Robert Wood Johnson Foundation (RWJF) Health & Society Scholars program at Harvard University and a RWJF seed grant (PI: Jennifer Karas Montez).The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the NIA, NINR, ORWH or the NIH.
Clinical Centers: University of Michigan, Ann Arbor – Siobán Harlow, PI 2011 – present, MaryFran Sowers, PI 1994–2011; Massachusetts General Hospital, Boston, MA – Joel Finkelstein, PI 1999 – present; Robert Neer, PI 1994 – 1999; Rush University, Rush University Medical Center, Chicago, IL – Howard Kravitz, PI 2009 – present; Lynda Powell, PI 1994 – 2009; University of California, Davis/Kaiser – Ellen Gold, PI; University of California, Los Angeles – Gail Greendale, PI; Albert Einstein College of Medicine, Bronx, NY – Carol Derby, PI 2011 – present, Rachel Wildman, PI 2010 – 2011; Nanette Santoro, PI 2004 – 2010; University of Medicine and Dentistry – New Jersey Medical School, Newark – Gerson Weiss, PI 1994 – 2004; and the University of Pittsburgh, Pittsburgh, PA – Karen Matthews, PI.
NIH Program Office: National Institute on Aging, Bethesda, MD – Winifred Rossi 2012 - present; Sherry Sherman 1994 – 2012; Marcia Ory 1994 – 2001; National Institute of Nursing Research, Bethesda, MD – Program Officers.
Central Laboratory: University of Michigan, Ann Arbor – Daniel McConnell (Central Ligand Assay Satellite Services).
Coordinating Center: University of Pittsburgh, Pittsburgh, PA – Maria Mori Brooks, PI 2012 - present; Kim Sutton-Tyrrell, PI 2001 – 2012; New England Research Institutes, Watertown, MA - Sonja McKinlay, PI 1995 – 2001.
Steering Committee: Susan Johnson, Current Chair; Chris Gallagher, Former Chair.
We thank the study staff at each site and all the women who participated in SWAN.
Abbreviations
- SES
Socioeconomic status
- CVD
Cardiovascular disease
- CRP
C-reactive protein
- PAI-1
Plasminogen activator inhibitor-1
- BMI
Body mass index
- SWAN
Study of Women’s Health Across the Nation
- CVs
Coefficients of variation
- ELISA
Enzyme-linked immunosorbent assay
Footnotes
Authors have declared no conflict of interest.
Reference List
- 1.Danesh J, Lewington S, Thompson SG, Lowe GD, Collins R, Kostis JB, Wilson AC, Folsom AR, Wu K, Benderly M, Goldbourt U, Willeit J, Kiechl S, Yarnell JW, Sweetnam PM, Elwood PC, Cushman M, Psaty BM, Tracy RP, Tybjaerg-Hansen A, Haverkate F, de Maat MP, Fowkes FG, Lee AJ, Smith FB, Salomaa V, Harald K, Rasi R, Vahtera E, Jousilahti P, Pekkanen J, D’Agostino R, Kannel WB, Wilson PW, Tofler G, Arocha-Pinango CL, Rodriguez-Larralde A, Nagy E, Mijares M, Espinosa R, Rodriquez-Roa E, Ryder E, Diez-Ewald MP, Campos G, Fernandez V, Torres E, Marchioli R, Valagussa F, Rosengren A, Wilhelmsen L, Lappas G, Eriksson H, Cremer P, Nagel D, Curb JD, Rodriguez B, Yano K, Salonen JT, Nyyssonen K, Tuomainen TP, Hedblad B, Lind P, Loewel H, Koenig W, Meade TW, Cooper JA, De SB, Knottenbelt C, Miller GJ, Cooper JA, Bauer KA, Rosenberg RD, Sato S, Kitamura A, Naito Y, Palosuo T, Ducimetiere P, Amouyel P, Arveiler D, Evans AE, Ferrieres J, Juhan-Vague I, Bingham A, Schulte H, Assmann G, Cantin B, Lamarche B, Despres JP, Dagenais GR, Tunstall-Pedoe H, Woodward M, Ben-Shlomo Y, Davey SG, Palmieri V, Yeh JL, Rudnicka A, Ridker P, Rodeghiero F, Tosetto A, Shepherd J, Ford I, Robertson M, Brunner E, Shipley M, Feskens EJ, Kromhout D, Dickinson A, Ireland B, Juzwishin K, Kaptoge S, Lewington S, Memon A, Sarwar N, Walker M, Wheeler J, White I, Wood A. Plasma fibrinogen level and the risk of major cardiovascular diseases and nonvascular mortality: an individual participant meta-analysis. JAMA. 2005;294(14):1799–809. doi: 10.1001/jama.294.14.1799. [DOI] [PubMed] [Google Scholar]
- 2.Libby P. Inflammation and cardiovascular disease mechanisms. Am J Clin Nutr. 2006;83(2):456S–60S. doi: 10.1093/ajcn/83.2.456S. [DOI] [PubMed] [Google Scholar]
- 3.Pepys MB, Hirschfield GM. C-reactive protein: a critical update. J Clin Invest. 2003;111(12):1805–12. doi: 10.1172/JCI18921. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Herrick S, Blanc-Brude O, Gray A, Laurent G. Fibrinogen. Int J Biochem Cell Biol. 1999;31(7):741–6. doi: 10.1016/s1357-2725(99)00032-1. [DOI] [PubMed] [Google Scholar]
- 5.Kohler HP, Grant PJ. Plasminogen-activator inhibitor type 1 and coronary artery disease. N Engl J Med. 2000;342(24):1792–801. doi: 10.1056/NEJM200006153422406. [DOI] [PubMed] [Google Scholar]
- 6.Donati MB, Iacoviello L. Fibrinogen and factor VIIc levels: independent risk factors or markers of coronary disease risk? J Thromb Haemost. 2007;5(3):458–60. doi: 10.1111/j.1538-7836.2007.02400.x. [DOI] [PubMed] [Google Scholar]
- 7.Galobardes B, Smith GD, Lynch JW. Systematic review of the influence of childhood socioeconomic circumstances on risk for cardiovascular disease in adulthood. Ann Epidemiol. 2006;16(2):91–104. doi: 10.1016/j.annepidem.2005.06.053. [DOI] [PubMed] [Google Scholar]
- 8.Danese A, Moffitt TE, Harrington H, Milne BJ, Polanczyk G, Pariante CM, Poulton R, Caspi A. Adverse childhood experiences and adult risk factors for age-related disease. Depression, inflammation, and clustering of metabolic risk markers. Archives of Pediatric and Adolescent Medicine. 2009;163(12):1135–43. doi: 10.1001/archpediatrics.2009.214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hagger-Johnson GE, Shickle DA, Deary IJ, Roberts BA. Direct and indirect pathways connecting cognitive ability with cardiovascular disease risk: socioeconomic status and multiple health behaviors. Psychosom Med. 2010;72(8):777–85. doi: 10.1097/PSY.0b013e3181ebf064. [DOI] [PubMed] [Google Scholar]
- 10.Pollitt RA, Kaufman JS, Rose KM, Diez-Roux AV, Zeng D, Heiss G. Early-life and adult socioeconomic status and inflammatory risk markers in adulthood. Eur J Epidemiol. 2007;22(1):55–66. doi: 10.1007/s10654-006-9082-1. [DOI] [PubMed] [Google Scholar]
- 11.Phillips JE, Marsland AL, Flory JD, Muldoon MF, Cohen S, Manuck SB. Parental education is related to C-reactive protein among female middle-aged community volunteers. Brain Behav Immun. 2009;23(5):677–83. doi: 10.1016/j.bbi.2009.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Camelo LV, Giatti L, Neves JA, Lotufo PA, Bensenor IM, Chor D, Griep RH, da Fonseca MJ, Vidigal PG, Kawachi I, Schmidt MI, Barreto SM. Life course socioeconomic position and C-reactive protein: mediating role of health-risk behaviors and metabolic alterations. The Brazilian Longitudinal Study of Adult Health (ELSA-Brasil) PLoS One. 2014;9(10):e108426. doi: 10.1371/journal.pone.0108426. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Nazmi A, Oliveira IO, Horta BL, Gigante DP, Victora CG. Lifecourse socioeconomic trajectories and C-reactive protein levels in young adults: findings from a Brazilian birth cohort. Soc Sci Med. 2010;70(8):1229–36. doi: 10.1016/j.socscimed.2009.12.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Loucks EB, Pilote L, Lynch JW, Richard H, Almeida ND, Benjamin EJ, Murabito JM. Life course socioeconomic position is associated with inflammatory markers: the Framingham Offspring Study. Soc Sci Med. 2010;71(1):187–95. doi: 10.1016/j.socscimed.2010.03.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kivimaki M, Lawlor DA, Juonala M, Smith GD, Elovainio M, Keltikangas-Jarvinen L, Vahtera J, Viikari JS, Raitakari OT. Lifecourse socioeconomic position, C-reactive protein, and carotid intima-media thickness in young adults: the cardiovascular risk in Young Finns Study. Arterioscler Thromb Vasc Biol. 2005;25(10):2197–202. doi: 10.1161/01.ATV.0000183729.91449.6e. [DOI] [PubMed] [Google Scholar]
- 16.Ridker PM, Hennekens CH, Rifai N, Buring JE, Manson JE. Hormone replacement therapy and increased plasma concentration of C-reactive protein. Circulation. 1999;100(7):713–6. doi: 10.1161/01.cir.100.7.713. [DOI] [PubMed] [Google Scholar]
- 17.Lakoski SG, Cushman M, Criqui M, Rundek T, Blumenthal RS, D’Agostino RB, Jr, Herrington DM. Gender and C-reactive protein: data from the Multiethnic Study of Atherosclerosis (MESA) cohort. Am Heart J. 2006;152(3):593–8. doi: 10.1016/j.ahj.2006.02.015. [DOI] [PubMed] [Google Scholar]
- 18.Cartier A, Cote M, Lemieux I, Perusse L, Tremblay A, Bouchard C, Despres JP. Sex differences in inflammatory markers: what is the contribution of visceral adiposity? Am J Clin Nutr. 2009;89(5):1307–14. doi: 10.3945/ajcn.2008.27030. [DOI] [PubMed] [Google Scholar]
- 19.Sowers M, Crawford S, Sternfeld B, Morganstein D, Gold E, Greendale G, Evans D, Neer R, Matthews K, Sherman S, Lo A, Weiss G, Kelsey J. SWAN: a multi-center, multi-ethnic community-based cohort study of women and the menopausal transition. In: Lobo R, Marcus R, Kelsey J, editors. Menopause:Biology and Pathology. New York: Academic Press; 2000. [Google Scholar]
- 20.Muthen LK, Muthen BO. Mplus [Computer software] Los Angeles, CA: 2012. [Google Scholar]
- 21.Wang J, Wang X. Structural Equation Modeling. West Sussex: John Wiley & Sons Ltd; 2012. [Google Scholar]
- 22.Clark SL, Muthen BO. [Accessed 7–27–2015];Relating latent class analysis results to variables not included in the analysis. 2009 https://www.statmodel.com/download/relatinglca.pdf.
- 23.Nagin DS. Group-based modeling of development. London: Harvard University Press; 2005. [Google Scholar]
- 24.Iacobucci D. Mediation analysis and categorical variables: The final frontier. J Consum Psychol. 2012;22(4):582–94. doi: 10.1016/j.jcps.2012.03.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Sobel ME. Asymptotic confidence intervals for indirect effects in structural equation models. In: Leinhardt S, editor. Sociological Methodology 1982. Washington, DC: American Sociological Association; 1982. pp. 290–312. [Google Scholar]
- 26.Preacher KJ, Hayes AF. Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behav Res Methods. 2008;40(3):879–91. doi: 10.3758/brm.40.3.879. [DOI] [PubMed] [Google Scholar]
- 27.Batty GD, Lawlor DA, Macintyre S, Clark H, Leon DA. Accuracy of adults’ recall of childhood social class: findings from the Aberdeen children of the 1950s study. J Epidemiol Community Health. 2005;59(10):898–903. doi: 10.1136/jech.2004.030932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Krieger N, Okamoto A, Selby JV. Adult female twins’ recall of childhood social class and father’s education: a validation study for public health research. Am J Epidemiol. 1998;147(7):704–8. doi: 10.1093/oxfordjournals.aje.a009512. [DOI] [PubMed] [Google Scholar]
- 29.Slopen N, Koenen KC, Kubzansky LD. Childhood adversity and immune and inflammatory biomarkers associated with cardiovascular risk in youth: a systematic review. Brain Behav Immun. 2012;26(2):239–50. doi: 10.1016/j.bbi.2011.11.003. [DOI] [PubMed] [Google Scholar]
- 30.Matthews KA, Chang YF, Thurston RC, Bromberger JT. Child abuse is related to inflammation in mid-life women: role of obesity. Brain Behav Immun. 2014;36:29–34. doi: 10.1016/j.bbi.2013.09.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Danese A, Pariante CM, Caspi A, Taylor A, Poulton R. Childhood maltreatment predicts adult inflammation in a life-course study. Proc Natl Acad Sci U S A. 2007;104(4):1319–24. doi: 10.1073/pnas.0610362104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Danese A, Tan M. Childhood maltreatment and obesity: systematic review and meta-analysis. Mol Psychiatry. 2013 doi: 10.1038/mp.2013.54. [DOI] [PubMed] [Google Scholar]
- 33.Midei AJ, Matthews KA. Interpersonal violence in childhood as a risk factor for obesity: a systematic review of the literature and proposed pathways. Obes Rev. 2011;12(5):e159–e172. doi: 10.1111/j.1467-789X.2010.00823.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.