| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Heller School for Social Policy and Management, Brandeis University, Waltham, Massachusetts (E.G.); Harold and Margaret Milliken Hatch Laboratory of Neuroendocrinology, Rockefeller University, New York, NY (B.S.M.); the Center for Epidemiology and Biostatistics (B.H.) and the Division of Endocrinology (L.M.D.), Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio; and the Department of Psychiatry, University of California School of Medicine, San Francisco, California (N.E.A.).
Address correspondence and reprint requests to Elizabeth Goodman, MD, Heller School for Social Policy and Management, Brandeis University MS 35, 415 South Street, Waltham, MA 02453-9110. E-mail: goodman{at}brandeis.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: Non-Hispanic black and white high school students (N = 758) in a suburban Midwestern public school district had a physical examination to measure height, weight, and waist circumference and a fasting morning blood sample drawn to assess cortisol, insulin, glucose, glycosylated hemoglobin, fibrinogen, low-density lipoprotein (LDL) and high-density lipoprotein (HDL) cholesterol, and triglycerides. A cumulative risk score was created from these physiological measures and waist circumference. Information on parent education and household income was obtained from a parent in a separate survey. Generalized estimating equation models were used to assess the association of parent education to the risks and the cumulative risk score adjusting for age, gender, and race.
Results: Lower parent education was associated with higher insulin, higher glucose, greater insulin resistance, higher LDL cholesterol, lower HDL cholesterol, higher waist circumference, and higher body mass index (p <.05 for all), but not cortisol, fibrinogen, glycosylated hemoglobin, or triglycerides in adjusted analyses. Cumulative risk scores ranged from 0 to 7 and were highly skewed; the median risk score was 1. A total of 7.4% had risk scores of 4 or more. Lower parent education was also associated with higher cumulative risk score (p <.001) and this association was maintained after adjustment for body mass index. Risk scores were highest, on average, among those with insulin levels greater than 1 standard deviation above the mean (mean risk score = 3.2, standard error = 0.18, median = 3).
Conclusion: Lower parent education is associated with multiple metabolic risks and cumulative risk in adolescents, suggesting that there is a strong intergenerational transfer of education’s influence on cardiovascular health. Our data imply that regulation of insulin may be a key factor underlying the influence of lower parent education on cardiovascular health early in the life course.
Key Words: adolescence • socioeconomic status • disparities • insulin • allostatic load
Abbreviations: BMI = body mass index; C = cholesterol; LDL = low-density lipoprotein; HDL = high-density lipoprotein; trig = triglycerides; fibrin = fibrinogen; HgbA1c = glycosylated hemoglobin; HOMA = homeostasis model assessment; SES = socioeconomic status; IQ = interquartile; CVD = cardiovascular disease; CV = coefficient of variation; educ = education.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Over the past 2 decades, both obesity and diabetes have been increasing in all population subgroups, including adolescents (6–10). Type 2 diabetes, once thought to be an adult disease, is increasingly diagnosed among teens (7). At the same time as these secular trends in obesity and type 2 diabetes developed, disparities in SES have increased (11). The temporal association among increasing socioeconomic inequality, obesity, and diabetes may hold the clue to regulatory systems and potential biologic mediators of social status’ impact on health.
Hyperinsulinemia and concomitant insulin resistance are found in type 2 diabetics and the obese (12). Insulin is sensitive to signals along the hypothalamic–pituitary–adrenal (HPA) axis (13,14). Evidence suggests that lower social status causes dysregulation of the HPA axis (14–16). The resultant physiological changes may, over time, cause development and progression of CVD among those of lower social status (14,16,17). This heuristic approach identifies the HPA axis, a central regulatory system involved in the creation of social inequalities in health, with cortisol and insulin as primary hormonal factors influencing carbohydrate metabolism, inflammation, cholesterol metabolism, and adiposity. Interestingly, the clustering of these secondary outcomes has been well described in metabolic syndrome, another growing health concern among adolescents (18).
Adolescence is an important developmental period to study in relation to the association among social status, the HPA axis, and these cardiovascular risks for a number of reasons. First, there are physiological changes associated with pubertal development that reflect alterations in central regulatory systems such as the HPA axis. The known rise in insulin resistance during puberty is 1 example (19). Second, the ability to think abstractly develops during adolescence. This cognitive change may lead to alterations in how the social environment is perceived and, therefore, potentially increase stress and HPA axis activation. Third, there is clear evidence that asymptomatic CVD begins very early in life and is linked to psychologic, humoral, and anthropomorphic cardiovascular risk factors (20–25). SES gradients in some of these risks have been identified in youth (2,26,27). Additionally, both observational and longitudinal studies indicate that cardiovascular risk factor clustering among adolescents tracks into adulthood (21,28).
The purpose of this study was to explore among adolescents the associations between social status, specifically lower parent education, and biomarkers of cardiovascular risk that reflect our heuristic frame (14,16). We hypothesized that, among adolescents, lower parent education would be associated with elevated levels of cortisol, insulin, glucose, blood lipids, and fibrinogen, and with greater body mass index (BMI) and central adiposity. We also hypothesized that parent education would be associated with a cumulative risk score derived by combining these physiological risks. Such a cumulative risk score would provide a measure of the multiple metabolic pathways involved in the brain–body link that controls physiological adaptation to the social environment (29).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Study Procedures
All study procedures occurred between 7 and 11:30 am at the high school. The vast majority of subjects were seen before 10 am. A blood sample was drawn after a minimum 10-hour fast, which was verified before the venipuncture. The blood sample included 2 10-cc plasma-EDTA Vacutainer tubes and 1 5-cc serum separator tube. These were placed on ice and transported to the Children’s Hospital as soon as data collection was complete, where they were processed. In addition, all subjects had a physical examination to measure height, weight, and waist circumference according to specified protocols (31).
Measures
Parent Education
Because income changes from year to year (32) but parent education is likely to be static for teens, these analyses focus on the influence of parent education. A parent provided information on parent education through a separate survey obtained during informed consent or by mail if not received at enrollment. Self and current spouse/partner education was reported and the higher of these used in analyses. Categories used in analyses were less than or equal to high school (N = 157, 20.7%), some college or technical/vocational training beyond high school (N = 203, 26.8%), college graduate (N = 218, 28.8%), and professional training beyond college (N = 180, 23.7%).
Household Income
The parental respondent also reported total pretax household income from all sources combined for the past 12 months on the parent survey in $25,000 increments ranging from less than $25,000 (N = 119, 15.7%) to $100,00 or higher (N = 138, 18.2%). For those missing information on income (11.6%), income was imputed based on the parent’s reported educational level and the race of the student (31). All analyses using income adjusted for household size.
Demographic Covariates
Date of birth, used in age calculation, and gender were available from school demographic data. Parent-identified race/ethnicity of the student was available from the school district. Family structure (single vs. 2-parent home) and household size were derived from the student survey.
Biomarkers of Cardiovascular Risk
Cortisol was measured with a double antibody radioimmunoassay using rabbit anticortisol (ICN) as the primary antibody and goat antirabbit IgG (Antibodies, Inc.) as the second antibody. The tracer was 125I-cortisol. Intraassay coefficient of variation (CV) was approximately 5% and interassay CV was 8%. Serum cortisol, rather than salivary, was obtained because a venipuncture was occurring as part of the study procedures.
Plasma insulin concentration was measured by radioimmunoassay (RIA) using an antiinsulin serum raised in guinea pigs, 125I- labeled insulin as a standard, and a double-antibody method to separate bound from free tracer. The sensitivity is 2 pM, with intra- and interassay CVs of 5% and 8%, respectively.
Glucose was measured by an enzymatic method. Intra- and interassay CVs are 1.2% and 1.6%, respectively.
Insulin resistance results from the glucose and insulin assays were used to derive insulin resistance measured by the homeostasis model assessment scores (HOMA) model (33).
Glycosylated hemoglobin (HgbA1c) was measured using high pressure liquid chromatography with a Waters 2690 Separation Module and dual-wavelength absorbance detector Waters model 2487. The coefficient of variation was 3.9% to 5.0%.
Fibrinogen was measured with the Sysmex CA6000 coagulation analyzer with Dade Behring Thrombin Reagent (Dade Behring, Marburg, Germany). The interassay coefficient of variation on the Sysmex CA6000 is 4%. The detection range is 50 to 950 mg/dL.
Blood lipids cholesterol was measured using the Cholesterol/HP kit from Roche (Boehringer Mannheim). The intraassay coefficient of variation was 1.0% and the interassay coefficient of variation was 2.2%. For high-density lipoprotein cholesterol (HDL-C), the HDL C-plus kit from Roche was used. This is a direct measurement (rather than a precipitation method). The intraassay coefficient of variation was 1.3% and the interassay coefficient of variation is 2.6%. Lipid profiles were performed on the Hitachi 704. National Cholesterol Education Program performance criteria for accuracy and precision are followed. Triglycerides were measured using a single reagent system from Roche-BMD. The CV was 4%. Low-density lipoprotein cholesterol (LDL-C) was calculated according to the Friedwald equation (total cholesterol–HDL–triglycerides/5), except when triglycerides were above 350 mg/dL. In those 2 cases, a direct measurement using the Roche LDL-C plus reagent was made.
Body mass index (BMI, kg/m2) was calculated from measured height and weight as previously described (31).
Waist circumference, an index of central obesity, was measured using a fiberglass tape crossing over the umbilicus and the superior iliac crests. The mean of 2 measurements made at the end of a normal expiration was used in analyses.
Cumulative Risk Score
A cumulative risk score was developed from 9 of the biomarker measures. Because BMI and waist circumference were very strongly correlated (Spearman’s rho = 0.90, p <.001) and because waist circumference may better reflect metabolically active fat, waist circumference was included in the risk score whereas BMI was not. Thus, included factors were insulin, cortisol, glucose, glycosylated hemoglobin, fibrinogen, LDL-cholesterol, HDL-C, triglycerides, and waist circumference. Cut points for each risk were defined as greater than 1 standard deviation (SD) above the mean. Insulin resistance was not included, because it was redundant given the included factors of glucose and insulin.
Data Analyses
Nonparametric tests were used to assess bivariate relationships. Correlational analyses assessed the association of parent education to the measured biomarkers. Regression analyses then were used to adjust for the influence of age, gender, and race. Insulin, HOMA, and triglycerides were log transformed to reduce skewness in all regressions. We also used regression analyses to test whether the addition of household income, which was highly correlated with parent education (Spearman’s rho = 0.60, p <.001), increased model fit, as evidenced by a significant change in the F statistic. Such a change would indicate that income added explanatory power to our models. Last, to assess whether the relationships between parent education and other cardiovascular risks were independent of general weight status, we further adjusted for BMI the regression equations assessing the relationship of parent education to the other biomarkers. We used generalized estimating equations (GEEs) to perform regression analyses to account for the presence of siblings within 11.6% (N = 79) of the 676 participating families. GEE models provide a robust method for estimation of regression model parameters when dealing with correlated data, which was created by the common biologic, social, and genetic background shared by siblings in the study. We tested for interactions among race, gender, and parent education in these models. None were found. For multivariate analyses of the cumulative risk score, we performed ordinal logistic regression modeling assuming proportional cumulative risk.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
Association of Parent Education to Markers of Physiological Risk
Table 2 presents the correlations of parent education and household income to each of the measured biomarkers. Lower parent education was weakly associated with higher insulin, higher glucose, greater insulin resistance, higher fibrinogen, lower HDL C, higher waist circumference, and higher BMI. No associations were seen with cortisol, glycosylated hemoglobin, LDL-C, or triglycerides. In general, correlations between income and these biomarkers were lower than those with than parent education. The exception here was glycosylated hemoglobin, which was slightly more strongly correlated with income.
|
These bivariate associations were maintained in regression models that adjusted for age, gender, and race for all biomarkers but fibrinogen (Table 3). In addition, in adjusted analyses, a significant association between parent education and LDL cholesterol was revealed. With further adjustment for income, the relationship between parent education and glucose (p = .34), insulin (p = .07), and insulin resistance (p = .08) became nonsignificant. This suggests that the effect of education is not independent of household income in regard to these 3 factors, although, given the correlation of income and education, the low p value in relation to insulin and insulin resistance suggests that education may be an important factor vis-à-vis these biomarkers. Parent education remained independently associated with HDL-C, LDL-C, BMI, and waist circumference with adjustment for income. Additionally, adjustment for income did not increase model fit in any of these models but HDL-C (ßincome = 0.72, standard error [SE] = 0.29, p = .01).
|
We next adjusted the relationships between parent education and the other biomarkers for BMI to determine if the effect of parent education was independent of general weight status. These data are also shown in Table 3. BMI was strongly associated with all risk factors but cortisol. Adjustment for BMI caused the association between parent education and insulin, insulin resistance, and waist circumference to become nonsignificant, suggesting that adiposity may be the underlying factor explaining these associations. However, the relationships between parent education and glucose, LDL-C, and HDL-C remained significant, suggesting other mechanisms besides adiposity are involved in linking parent education to these physiological measures.
Cumulative Risk Score
The cumulative risk score ranged from 0 to 7 out of a possible 9 (Figure 1). Median score was 1. Mean score was 1.21 and SD 1.37. Nearly one fifth (17.4%) had risk scores of 3 or more. Risk scores were highest, on average, among those with insulin levels greater than 1 SD above the mean (mean risk score = 3.2, SE = 0.18, median = 3). Risks were widely distributed with no consistent pattern of correlation. Figure 2, which documents the distribution of risk factors among those with a given risk, illustrates this. There were no significant race, gender, or age differences in the cumulative risk score. However, lower parent education was significantly associated with higher cumulative risk score in ordinal logistic regression analyses (Table 4), and this relationship was independent of BMI. Of note, when adjustment for BMI as well as parent education was made, a significant relationship between race and the cumulative risk score was demonstrated. Black adolescents were more likely to have lower cumulative risk scores, which may reflect the better lipid profiles (higher HDL-C and lower triglycerides) noted among black participants. There were no significant race-by-SES interactions.
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This is one of the first studies to assess the relationship between social status and insulin among adolescents. Among adults, studies have demonstrated SES gradients in diseases associated with chronic insulin dysregulation (5,34,35) as well as between lower SES and insulin resistance (36). Gower et al. recently demonstrated an association between lower SES and the acute insulin response to glucose, but not fasting insulin, in a small study of mostly young, non-Hispanic white girls (37). In contrast, our study, which has equal representation of adolescent boys and girls who are evenly distributed between racial groups, demonstrated a significant association between lower parent education and both fasting insulin and insulin resistance.
Although we demonstrated an association between parent education and insulin, we were not able to show an association between social status and cortisol, the other potential hormonal mediator we measured. The association between cortisol and social status has been inconsistent in other studies (38,39). Whether this represents a developmental trend, low statistical power, or problems with the measurement of cortisol is unknown. We could not adjust for time of venipuncture or time of awakening. Thus, the diurnal pattern of cortisol secretion may have influenced our findings.
Measurement of cortisol, although problematic, was important in this study because 1 of the pathways through which social status is theorized to influence health is through physiological changes required to respond to the stress of living with chronic social disadvantage (16). The other major pathways are through behavioral and genetic effects (40). We could not assess genetic or behavioral effects in this study. However, we had an indirect measure of genetics and behaviors related to cardiovascular risk in the form of BMI. Diet and physical activity/inactivity are major behavioral determinants of BMI and influence many of these risks. In addition, genetic determinants of obesity are captured, albeit grossly, through this measure. The strong relationship between BMI and the other biomarkers and the effect of adjustment for BMI in our regressions support the importance of behaviors in relation to shaping health and the need to increase our understanding of genetic vulnerabilities (40). However, we also found that the relationship of parent education to the cumulative risk score, as well as to glucose, HDL-C, and LDL-C, was independent of adjustment for BMI. These exciting findings, especially the finding in relation to cumulative risk, support the idea that the social environment can influence physiological functioning directly rather than only through health behaviors.
One model that incorporated direct physiological effects, behaviors, and genetics is allostatic load (16). Allostatic load grew out of the concept of allostasis, which refers to the neurohormonal regulatory processes an organism uses to maintain its internal milieu in the face of changing environmental contexts (16,41). In a persistently challenging environment such as that created by low social status, chronic stress, in conjunction with lifestyle changes and genetic predispositions, contributes to dysregulation of multiple major regulatory pathways required for adaptation to environmental demands. The resulting cumulative burden on the body, or "allostatic load," has been linked to both acute illness and development of chronic diseases (42–44). Although the allostatic load concept is a powerful organizing theory for understanding the development of social inequalities in health, few studies apply the concept to children and adolescents, and most that do are cross-sectional and assess a single marker of physiological arousal rather than multiple measures (38,39,45). Our focus on cumulative risk and multiple metabolic pathways and our findings are consistent with the allostatic load hypothesis, although we lack long-term longitudinal data that would be needed to directly test this theory.
Some other limitations to this study should be noted. As a cross-sectional study, we cannot assess causality, although our independent variable, parent education, was likely present from birth for these youth. Because our cohort reflects the demographics of the Greater Cincinnati area, we did not have representation of other racial/ethnic groups besides non-Hispanic black and white youth. We were not able to assess functioning of the autonomic nervous system or immune system in this epidemiologic school-based study. These systems have been identified as other important regulatory systems that mediate social status’ influence on cardiovascular health. We do, however, have important strengths, including a large, socioeconomically diverse, gender-balanced cohort, parental report of parent education, and multiple biomarkers of risk.
Our findings suggest that insulin regulation could be an important mechanism through which lower social status creates cardiovascular health disparities. Such a hypothesis has been suggested in relation to metabolic syndrome and social inequalities in adult cardiovascular disease (35). Prevalence of metabolic syndrome in adolescence has been estimated to be between 4.2% and 8.4%, depending on the definition used (18). Although the etiology of metabolic syndrome is debatable, hyperinsulinemia is considered 1 of its core components, and by definition, multiple metabolic pathways are involved (46,47). The biomarkers assessed here relate to these same metabolic pathways. The major difference between our approach assessing cumulative risk and one that focuses on a particular clinical outcome such as metabolic syndrome is that our approach expands and highlights the multisystem dysregulation. A metabolic syndrome approach reduces such multiple pathway effects to a single, dichotomous end point. Although effective for identifying those at highest risk for clinical disease in older populations, this approach does not easily translate into primary prevention for children and adolescents because diseases such as atherosclerotic heart disease are multifactorial, have a long induction period, and will likely not be manifest until adulthood. Additionally, information on the natural history of the metabolic dysregulation may be lost with such a strategy. These data suggest that developmental models of disease risk may provide a more comprehensive theoretical base for studies of life-course trajectories in health. Our findings indicate that lower social status influences multiple metabolic pathways early in life, even in apparently healthy individuals. How and if this dysregulation across systems leads to health disparities in adulthood will require further longitudinal research.
The authors thank Stephen R. Daniels for his advice and support, as well as the students, parents, teachers, administration, and staff of the Princeton City School district and the PSD study staff.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This work was supported by NIH grants HD41527, DK59183, and M01 RR 08084, the William T. Grant Foundation, and the John D. and Catherine T. MacArthur Foundation. This work was presented, in part, at the North American Association for the Study of Obesity’s Annual meeting, Las Vegas, NV, November 17, 2004.
DOI:10.1097/01.psy.0000149254.36133.1a
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. Sanders-Phillips, B. Settles-Reaves, D. Walker, and J. Brownlow Social Inequality and Racial Discrimination: Risk Factors for Health Disparities in Children of Color Pediatrics, November 1, 2009; 124(Supplement_3): S176 - S186. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. B Dowd, A. M Simanek, and A. E Aiello Socio-economic status, cortisol and allostatic load: a review of the literature Int. J. Epidemiol., October 1, 2009; 38(5): 1297 - 1309. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. H. Stea, M. Wandel, M. A. Mansoor, S. Uglem, and W. Frolich BMI, lipid profile, physical fitness and smoking habits of young male adults and the association with parental education Eur J Public Health, January 1, 2009; 19(1): 46 - 51. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C Power, K Atherton, and O Manor Co-occurrence of risk factors for cardiovascular disease by social class: 1958 British birth cohort J Epidemiol Community Health, December 1, 2008; 62(12): 1030 - 1035. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Goodman, S. R. Daniels, J. B. Meigs, and L. M. Dolan Instability in the Diagnosis of Metabolic Syndrome in Adolescents Circulation, May 1, 2007; 115(17): 2316 - 2322. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. D. Hanson and E. Chen Socioeconomic Status, Race, and Body Mass Index: The Mediating Role of Physical Activity and Sedentary Behaviors during Adolescence J. Pediatr. Psychol., April 1, 2007; 32(3): 250 - 259. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Goodman, S. R. Daniels, and L. M. Dolan Socioeconomic Disparities in Insulin Resistance: Results From the Princeton School District Study Psychosom Med, January 1, 2007; 69(1): 61 - 67. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C R Chittleborough, F E Baum, A W Taylor, and J E Hiller A life-course approach to measuring socioeconomic position in population health surveillance systems. J Epidemiol Community Health, November 1, 2006; 60(11): 981 - 992. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
