• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Soc Sci Med. Author manuscript; available in PMC Jan 1, 2012.
Published in final edited form as:
PMCID: PMC3039215

Do biological measures mediate the relationship between education and health: a comparative study

Noreen Goldman, D.Sc., Cassio Turra, Ph.D., Luis Rosero-Bixby, Ph.D., David Weir, Ph.D., and Eileen Crimmins, Ph.D.


Despite a myriad of studies examining the relationship between socioeconomic status and health outcomes, few have assessed the extent to which biological markers of chronic disease account for social disparities in health. Studies that have examined this issue have generally been based on surveys in wealthy countries that include a small set of clinical markers of cardiovascular disease. The availability of recent data from nationally representative surveys of older adults in Costa Rica and Taiwan that collected a rich set of biomarkers comparable to those in a recent US survey permits us to explore these associations across diverse populations. Similar regression models were estimated on three data sets – the Social Environment and Biomarkers of Aging Study in Taiwan, the Costa Rican Study on Longevity and Healthy Aging, and the Health and Retirement Study in the USA – in order to assess (1) the strength of the associations between educational attainment and a broad range of biomarkers; and (2) the extent to which these biomarkers account for the relationships between education and two measures of health status (self-rated health, functional limitations) in older populations. The estimates suggest non-systematic and weak associations between education and high risk biomarker values in Taiwan and Costa Rica, in contrast to generally negative and significant associations in the US, especially among women. The results also reveal negligible or modest contributions of the biomarkers to educational disparities in the health outcomes. The findings are generally consistent with previous research suggesting stronger associations between socioeconomic status and health in wealthy countries than in middle income countries and may reflect higher levels of social stratification in the US. With access to an increasing number of longitudinal biosocial surveys, researchers may be better able to distinguish true variations in the relationship between socioeconomic status and health across different settings from methodological differences.

Keywords: Costa Rica, Taiwan, USA, biomarkers, biosocial survey, socioeconomic disparities, health outcomes, longitudinal


Despite a long-standing interest in social inequalities in health and survival, and in the behavioral, psychosocial, and environmental mechanisms that may account for these disparities, social scientists have only recently begun to examine the underlying biological pathways linking social position to mental and physical well-being. Interest in these physiological connections has led to a proliferation of “biosocial surveys” that obtain socio-demographic information through interviews along with biological markers related to chronic disease based on physical assessments and laboratory analyses (Weinstein, Vaupel, & Wachter 2008). These surveys are providing researchers with measurements of biomarkers related to metabolic and cardiovascular disease, often combined with indicators of immune and neuroendocrine function, for broad population-based samples. The resulting biological markers not only offer researchers more objective assessments of health status and disease than the self-reported information typically collected in household surveys, but they are likely to generate insights into the nature of the physiological dysregulation and, ultimately, the underlying causal pathways linking lower social status to poorer health.

Previous studies examining socioeconomic differentials in biomarkers or the impact of these differentials on social disparities in health have largely been based on data from wealthy Western nations. Moreover, many studies have been limited to a few biological markers derived from small or select samples. The availability of recent data from nationally representative surveys of older adults in two middle-income countries in different regions (Costa Rica and Taiwan) that collected a rich set of biological measures comparable to those in a recent US survey permits us to explore these relationships in a diverse set of populations.

These countries provide a fascinating set of comparisons. Despite substantial differences in levels of human development in the recent past, the US, Taiwan and Costa Rica now have almost identical life expectancies at birth (approximately 78 years in Taiwan and the US and 79 in Costa Rica; (Population Reference Bureau)). The average educational level is currently the same in Taiwan and the US (schooling life expectancy of 15.6 years in 2005), but several years lower in Costa Rica (11.7 years in 2005) (UNESCO Institute for Statistics 2009; United Nations Educational, Scientific and Cultural Organization). However, the countries continue to have substantially different levels of economic well-being, health care expenditures and inequities in health care. GDP per capita (PPP) in 2009 is much lower in Costa Rica ([congruent with]$11,000) than in either Taiwan ([congruent with] $33,000) or the US ([congruent with] $47000) (International Monetary Fund 2010). Moreover, in contrast to the decentralized health care system in the US that leaves a large fraction of the population without health insurance (until age 65) or adequate health services, Taiwan and Costa Rica have national health insurance systems that cover the vast majority of residents. This health care is achieved at a fraction of the cost of health care in the US: health expenditures per capita in the US are about six times as high as in Taiwan and about nine times as high as in Costa Rica (Lu & Hsiao 2003; Unger, De Paepe, Buitron, & Soors 2008).


A huge body of research has examined the relationships between socioeconomic status (SES) – most commonly measured by education, income, and occupational status – and health outcomes. There is a consensus that more educated and wealthier individuals fare better than their lower SES counterparts on most broad measures of health status and that these patterns persist at the older ages, although the differentials are generally smaller than at younger ages (Smith & Goldman 2007; Zajacova, Goldman, & Rodriguez 2009). However, the few studies conducted in middle-income countries suggest a more erratic and weaker relationship between measures of SES and health than in rich countries (Rosero-Bixby & Dow 2009; Smith & Goldman 2007).

Cardiovascular disease, typically the leading cause-of-death in Western populations, is a presumed source of many of the socio-economic differentials in health and survival. Many studies have examined the association between SES and clinical markers of cardiovascular and metabolic function (e.g., blood pressure, total and HDL cholesterol, body-mass index, and glucose levels). Relatively few population-based studies have examined other physiological markers related to stress and health, such as neuroendocrine measures (Love, Seeman, Weinstein, & Ryff 2010), even though exposure and response to stressful experience are presumed to comprise an important pathway linking lower social status to poorer health (Siegrist & Marmot 2004; Steptoe & Marmot 2002).

There is a pervasive notion that social inequalities in health are reflected in SES differentials in biomarkers related to chronic disease (see, for example, Banks, Marmot, Oldfield, & Smith 2006; Kristenson, Eriksen, Sluiter, Starke, & Ursin 2004), but a more nuanced assessment of findings suggests that the associations between biomarkers and SES are not clear-cut. For example, blood pressure which is one of the most commonly analyzed biomarkers, has an inverse association with SES in numerous studies (Atherton & Power 2007; Banks, Marmot, Oldfield, et al. 2006; Bobak, Hertzman, Skodova, & Marmot 1999; Martikainen, Ishizaki, Marmot, Nakagawa, & Kagamimori 2001), no significant association in many others (Brunner, Marmot, Nanchahal, et al. 1997; Kapuku, Treiber, & Davis 2002; Steptoe, Kunz-Ebrecht, Owen, et al. 2003) and a positive association in at least one study (Reddy, Rao, & Reddy 2002). Moreover, the findings are often not consistent within a given study, showing variations by sex or by type of indicator of SES. Additional evidence from a comparative study in Japan and England suggests that the relationships between SES and cardiovascular risk factors are likely to vary across cultural and socioeconomic settings (Martikainen, Ishizaki, Marmot, et al. 2001).

Recent studies in Costa Rica and in Taiwan underscore the non-systematic associations between SES and biomarkers of chronic disease and suggest that, as with SES disparities in health outcomes, SES differentials in biological risk factors may be weaker in moderate to low income countries than in wealthy nations (Dowd & Goldman 2006; Rosero-Bixby & Dow 2009). In this study, we take advantage of the availability of comparable high quality biological and health data from Costa Rica, Taiwan and the US to move beyond single-country analyses in order to assess (1) the strength and direction of the associations between education and a broad range of biomarkers; and (2) the extent to which these biomarkers account for the relationships between education and two measures of health status in older populations. Based on the previous work summarized above, we hypothesize that the associations between education and the biomarkers will be considerably weaker in Taiwan and Costa Rica than in the US, and, consequently, that only in the US are we likely to find that the biomarkers mediate the association between education and our downstream measures of health.

Materials and Methods


Data for this analysis come from three sources: the Social Environment and Biomarkers of Aging Study (SEBAS) in Taiwan, the Costa Rican Study on Longevity and Healthy Aging (CRELES) and the Health and Retirement Study (HRS) in the US. The three surveys comprise information regarding demographic and socioeconomic characteristics, physical health, health-related behaviors, psychological well-being and health service utilization. In addition, all surveys collected physiological data that provide a comparable set of biological measures. Prior to data collection, institutional review boards at the participating institutions approved all survey procedures and respondents provided written informed consent for participation in the interviews and biomarker collection.

SEBAS is based on a follow-up of the Survey of Health and Living Status of the Near Elderly and Elderly in Taiwan (TLSA), a nationally representative longitudinal survey (including the institutionalized population at baseline) that was administered six times between 1989 and 2007. In 2000, a random subsample of respondents for SEBAS was drawn from surviving respondents in 1999, with an oversampling of persons aged 70 years and older and persons in urban areas. A second wave of the survey was conducted in 2006, but the present analysis relies on data from the first wave. SEBAS, which includes respondents 54 and older at the 2000 interview, consists of a face-to-face in-home interview and a medical exam conducted at a hospital several weeks after the interview.

CRELES is an on-going longitudinal study of a nationally representative sample of adults born in 1945 or earlier (ages 60 and over at the first interview) and residing in Costa Rica, with over-sampling of the oldest old. For this analysis we use data for the first wave of interviews, conducted between 2004 and 2006, mostly in 2005. The interview data and biological specimens were collected in the participants’ homes, usually in two visits.

HRS started in 1992 as a nationally representative study of the non-institutionalized population aged 51 to 61 and their spouses/partners. Since 1998, when HRS was merged with the Study of Assets and Health Dynamics among the Oldest Old (AHEAD), HRS has surveyed a nationally representative sample of Americans over the age of 50 every two years. In the 2006 HRS wave, one-half of the full sample (ages 53 and over) was randomly selected to provide an enhanced face-to-face interview that included anthropometric, physical performance and biomarker measurements (Weir 2008). The data and specimens were collected at the participants’ homes. Table 1 provides a summary of the main characteristics of each survey.

Table 1
Summary of survey characteristics for SEBAS, CRELES and HRS


In order to preserve comparability across populations, we limit the analysis to biomarkers that were ascertained in at least two of the surveys (SEBAS and CRELES obtained more markers than HRS). Of the ten markers examined here, eight are related to metabolic syndrome and two are nonclinical measures associated with neuroendocrine functioning: urinary cortisol and dehydroepiandrosterone sulfate (DHEAS). We construct dichotomous measures of the biomarkers – coded as 1 when the respondent has a high risk value and 0 otherwise – to capture values outside established cutoff points of clinical markers as well as the potential for risk at extreme values of nonclinical markers (see Table 2 for cutoff points). Binary measures of biomarkers, which have been used extensively in other analyses (Feldman, Johannes, Araujo, Mohr, Longcope, & McKinlay 2001; Park, Cho, Song, & Sung 2006), have several advantages over continuous parameterizations. First, a linear specification is unsatisfactory for biomarkers that have risk associated with both low and high values; a non-monotonic specification such as a quadratic function can be readily implemented when biomarkers are explanatory – but not outcome – variables. Second, because most of the markers in this analysis have well-established cutoff points designating clinical risk, an underlying assumption of a threshold effect on health status is at least as plausible as a linear effect throughout the range of the biomarker. Third, research related to the construction of indices of physiological dysregulation has concluded that indices based on counts of biomarkers with high risk values (i.e., a sum of binary variables) perform as well as those based on continuous markers (Seplaki, Goldman, Glei, & Weinstein 2005). Finally, in contrast to continuous measures, binary variables are robust to outliers.

Table 2
Cutoff points and summary measures for high risk values of individual biomarkers in SEBAS, CRELES and HRS

The metabolic syndrome markers include two indicators of body fatness: waist circumference and BMI. For waist circumference, we code values larger than 88 centimeters for women and 102 for men as high risk (National Heart, Lung, and Blood Institute, Obesity Education Initiative). For BMI, calculated as weight divided by height squared (kg/m2), we recode the original values to reflect obesity (values larger than or equal to 30) and underweight status (lower than 18.5), following earlier studies that have shown that both excess fat and accelerated loss of lean body mass are associated with health deterioration (Allison, Faith, Heo, & Kotler 1997; Seidell & Visscher 2000). (Some studies recommend the use of lower cutoff points to define obesity in East Asian populations, but this is a controversial issue (World Health Organization 1998).) For systolic and diastolic blood pressure, values ≥ 140 and 90 mmHg respectively (based on two readings for SEBAS and CRELES and three readings for HRS) are defined as high risk (Whitworth 2003). We also include measures of total serum cholesterol (high risk 240 mg/dL) and triglycerides (high risk ≥ 200 mg/dL), based on fasting blood specimens in SEBAS and CRELES and a non-fasting blood specimen in the HRS (Third Report of the National Cholesterol Education Program (NCEP) Expert Panel 2002). Two biomarkers measure glucose metabolism: fasting glucose and glycosylated hemoglobin (HbA1c). High risk cutoff values for these measures are ≥ 100 mg/dl and > 6.5% respectively (AACE Diabetes Mellitus Clinical Practice Guidelines Task Force 2007).

In the absence of guidelines for normal ranges of nonclinical markers, we use cutoff points for cortisol and DHEAS based on the distributions of these biomarkers in each survey (Seeman, Singer, Rowe, Horwitz, & McEwen 1997), calculated separately for men and women. Because both low and high cortisol levels may be related to increased risk of dying (Marklund, Peltonen, Nilsson, & Olsson 2004), this marker equals 1 for respondents who have values in the lowest or highest deciles. DHEAS equals 1 only for individuals in the lowest quartile because low levels of DHEAS have been related to several disorders, including impairments of immune and cognitive function, as well as mortality (Barrett-Connor, Khaw, & Yen 1986; Mazat, Lafont, Berr, et al. 2001). Triglycerides, fasting glucose, cortisol and DHEAS measurements are not available in the HRS, and, thus, are compared only in SEBAS and CRELES.

The health outcomes in this analysis comprise two measures – self-rated health and functional limitations – that have been shown to be associated with a broad range of biomarkers (Goldman, Turra, Glei, Lin, & Weinstein 2006; Jylhä, Volpato, & Guralnik 2006; Koster, Penninx, Bosma, et al. 2005). Self-rated health is reported according to the conventional 5-point ordinal scale: excellent, very good, good, fair and poor. The measure of functional limitations is constructed to be comparable across surveys and reflects the number of activities that the respondent reports difficulty performing; this count is derived from four mobility tasks (lifting or carrying weight, raising arms above shoulders, walking a specified distance, and climbing stairs), two instrumental activities of daily living (buying personal items and managing money), and three activities of daily living (bathing, eating, and toileting).

We use educational attainment as our measure of SES for several reasons. Educational attainment is not only the most comparable measure across the three settings, but it is also better suited to measure SES at older ages than income and occupation and has less missing data than the other variables. Wealth has been found to be important in some studies of older populations (Avendano & Glymour 2008; Bond Huie, Krueger, Rogers, & Hummer 2003), but it is not consistently measured in the studies. In addition, education, which is typically completed early in life, is considerably less likely to be affected by poor health – i.e., reverse causality – than are measures of wealth, income and occupational status (Elo 2009). Although, we recognize that education is an imperfect measure of social position among older adults, particularly in modernizing societies such as Taiwan that have experienced dramatic increases in educational attainment and social mobility, there is evidence that older adults give more weight to their schooling than to other economic indicators when evaluating their social position (Goldman, Cornman, & Ming-Cheng Chang 2006). In order to preserve comparability, we code education into three categories based on approximate terciles of its (unweighted) distribution in each survey. The cutoff points are similar in Taiwan (men: 0–5, 6, 7+; women: 0, 1–6, 7+) and Costa Rica (men: 0–1, 2–5, 6+; women: 0–2, 3–5, 6+). In the US, where secondary education has been mandatory for a longer period, the cutoff points are 0–11 12, and 13+ years of schooling for both sexes. Distributions of educational attainment, as well as self-rated health and functional limitations, are provided in Supplementary Table 1[INSERT LINK TO ONLINE FILE]. We include linear and quadratic controls for age in all models.

Analytic Strategy

To examine the associations between education and the physiological measures in each of the three countries, we estimate separate logistic regression models for each biomarker, controlling for age. We fit separate models for men and women because of presumed sex differences in the biological mechanisms linking SES and health (Dowd & Goldman 2006; Rosero-Bixby & Dow 2009).

To test for the mediating effects of biomarkers in the relation between education and health, we also estimate separate models for each population and sex. Ordered logistic regression models are used for self-rated health and Poisson regression models for the count of functional limitations. We compare two models for each outcome. The first includes only age and education, whereas all biomarkers are included in the second model. The analytic sample sizes vary slightly across models for a given country, due primarily to missing values for biomarkers. We use Stata 10 to estimate the models (StataCorp 2008).


Table 2, which provides weighted estimates of the prevalence of high risk biomarker values, reveals considerable variation across the three countries. The frequencies of extreme BMI values and large waist circumference are higher in the U.S compared to Taiwan and Costa Rica. In contrast, Costa Rica has the largest proportions of persons with high blood pressure and high cholesterol. Estimates that include persons using antihypertensive medication regardless of their blood pressure levels (not shown) reveal that the US and Costa Rica have similarly high rates of hypertension under these revised definitions.

Estimated odds ratios for education from the logistic models of having high risk values of each biomarker are presented in Tables 3 and and4.4. In Taiwan, only one of the 10 markers is significantly associated with education for both sexes (DHEAS); glucose is significant only for men, whereas three other markers are significant only for women (BMI, diastolic blood pressure and cortisol). Even fewer significant associations are present in Costa Rica – glucose and triglyceride values among men, and systolic blood pressure, glucose, and glycosylated hemoglobin values among women – and the direction of the association with education varies across these markers.

Table 3
Odd ratios from logistic models of having high risk values of each biomarker, by education: men in SEBAS, CRELES and HRS
Table 4
Odds ratios from logistic models of having high risk values of each biomarker, by education: women in SEBAS, CRELES and HRS

In the US, where we have data on only six of the 10 markers, we find a higher proportion of statistically significant associations: among women, all associations except those for cholesterol are significant, and, for men, three of the biomarkers (diastolic and systolic blood pressure and HbA1c) have a significant odds ratio associated with education. Most of the significant associations in each country pertain to the highest education group, whose members are less likely than those in the lowest level to have high risk values (except for Costa Rican men). Only in the US do we generally find a graded relationship, with the odds ratios becoming progressively smaller for higher levels of education, particularly for women.

Despite only sporadic associations between education and the biomarkers for Costa Rica and Taiwan, education is significantly related to both self-rated health and functional limitations in all samples (Model 1, Tables 5 and and6).6). The generally larger coefficients (in absolute value) for HRS suggest that the relationship between higher levels of education and better health outcomes is stronger in the US than in Costa Rica or Taiwan, consistent with the literature identifying weaker associations in middle income countries. As shown by a comparison of estimates from Models 1 and 2, evidence supporting the mediating role of biological markers is modest. In most cases, the coefficients and significance levels for the education variables change little with the inclusion of the biomarkers, suggesting that the biomarkers do not help to explain the SES disparities in these health outcomes. However, for the most educated women in Taiwan and the US, and, to a lesser extent, educated Taiwanese men, the biomarkers appear to mediate part of the association between education and the health measures. Nevertheless, in the presence of controls for biomarkers, all but one of these associations remains statistically significant.

Table 5
Estimated regression coefficientsa for self-rated health and functional limitations, by education: men in SEBAS, CRELES and HRS
Table 6
Estimated regression coefficientsa for self-rated health and functional limitations, by education: women in SEBAS, CRELES and HRS


Overall, these results suggest non-systematic associations between education and high risk biomarker values in Taiwan and Costa Rica, and negligible or modest mediating effects of the biomarkers on educational disparities in two health measures for all three countries. Particularly in light of the fact that the statistical model controls for only six biomarkers in the US in contrast to 10 markers in the other countries, the findings suggest a larger contribution of the biomarkers to educational disparities in the US than in Taiwan and Costa Rica. In models not presented here, we explored the robustness of these findings by specifying educational attainment based on conventional schooling levels. The alternative formulations produced results similar to those described above. We also redefined high blood pressure and high levels of glycosylated hemoglobin to include respondents taking medications for these conditions. The corresponding estimates of prevalence increased under the revised definitions, but there was little effect on the education coefficients. Finally, we evaluated the robustness of our findings to the specification of the biomarkers. After trimming outlying values of the biomarkers, we (1) estimated linear regression models for each biomarker (corresponding to the logistic models in Tables 3 and and4)4) and (2) re-estimated Model 2 in Tables 5 and and66 based on linear, rather than binary, parameterizations of the markers. Overall, about the same number of biomarkers were significantly associated with education in the binary and linear specifications. The one exception was a larger number of significant associations for Costa Rican males based on linear markers, but all of these were in the same direction as those for the binary measures, suggesting that higher levels of education are related to riskier values of the markers. The comparison of education coefficients with and without controls for the biomarkers revealed negligible or modest changes under both specifications, but, overall, the reduction in the magnitude of the education coefficients was slightly larger with binary markers for Costa Rica and Taiwan and slightly larger with linear markers for the US.

Although the results for Taiwan and Costa Rica challenge the commonly held assumption that more educated individuals have healthier biological profiles than their less educated peers, the findings are consistent with earlier studies based on SEBAS and CRELES, each of which used a somewhat different set of markers, SES variables, and health outcomes than the present study. The Taiwan study found that biomarkers associated with the cardiovascular, neuroendocrine, and immune systems explained relatively little of the association between SES and health status, primarily because few biomarker distributions were significantly associated with education and income (Dowd & Goldman 2006). Although analyses of the Costa Rican data did not assess the extent to which biomarkers accounted for SES differentials, the results demonstrated that (1) the direction of the SES gradient varied across biomarkers, with lower SES individuals often having better indicators than their higher SES counterparts (Rosero-Bixby & Dow 2009); and (2) fewer cardiovascular risk factors were associated with education in Costa Rica than in the US, based on NHANES (Rehkopf, Dow, & Rosero Bixby 2010).

The findings for HRS are generally consistent with those in other US studies. For example, analyses based on NHANES indicate that high income and high education levels are negatively associated with high blood pressure and total cholesterol and positively associated with HDL cholesterol, as expected (Kanjilal, Gregg, Cheng, et al. 2006; Muennig, Sohler, & Mahato 2007). However, few of these cardiovascular markers are significantly related to education in an analysis based on the MacArthur Successful Aging Study (Seeman, Crimmins, Huang, et al. 2004), perhaps because the MacArthur sample is older (ages 70–79) than the sample in HRS and in most other US studies. The presence of significant inverse associations between SES and markers of obesity among American women but not American men – as found here – has been identified in studies based on NHANES (Chang & Lauderdale 2005; Zhang & Wang 2004).

In contrast to a fairly large literature examining links between SES and biomarkers, few studies in the US have looked at the mediating effects of a substantial set of biomarkers (i.e., more than several markers or markers pertaining to more than one physiological system) on SES differentials in health. An analysis based on the MacArthur Study identifies modest effects of cardiovascular and immune parameters, and very small effects of neuroendocrine markers, on the relative risks of dying associated with educational attainment (Seeman, Crimmins, Huang, et al. 2004). Findings for other countries vary substantially by time and place. For example, a study in Eastern Finland shows large reductions in the relative hazards of all-cause mortality after adjustment for biologic risk factors (Lynch, Kaplan, Cohen, Tuomilehto, & Salonen 1996). Yet, another Finnish study demonstrates that, since the 1980s, traditional risk factors have been less able to account for socioeconomic disparities in cardiovascular mortality in Finland as compared with earlier periods, perhaps because of improved access to new medical technologies (Harald, Koskinen, Jousilahti, Torppa, Vartiainen, & Salomaa 2008). In London, an analysis based on the Whitehall II study finds large effects of metabolic and inflammatory markers on differentials in incident coronary disease by employment grade (Marmot, Shipley, Hemingway, Head, & Brunner 2008). In contrast, two studies in South Korea fail to find mediating effects of biologic risk factors on either cardiovascular disease or all-cause mortality (Khang & Kim 2005; Yun-Mi Song, Ferrer, Sung-il Cho, Sung, Ebrahim, & Smith 2006).

Despite a justified appeal for international comparisons of social gradients in health that integrate biological mechanisms (Banks, Marmot, Oldfield, et al. 2006; Elo 2009), such undertakings are generally unable to establish whether divergent findings reflect true variability in the physiological pathways linking SES to health across countries, regions and time periods; differences across data sets in measurement error or definitions of biomarkers, SES, and health outcomes; differences in analytic strategies; or variations in sample size. In this analysis, measurement error, particularly for the biomarkers – which are collected on a single day – is a serious concern. An additional limitation is the disparity in sample sizes (HRS is the largest and SEBAS the smallest of the surveys) and associated levels of statistical power. In particular, the finding that there are more significant associations between biomarkers and education for the US as compared to Taiwan and Costa Rica could reflect, in part, the larger sample size of the HRS, although sample size considerations would not account for Costa Rica displaying weaker associations than Taiwan. Nevertheless, the similarity in the variables, statistical models, and analytic strategy for the three data sets strengthens our conclusion that the biomarkers are most strongly associated with education in the US and that their mediating effects on health are likely to be larger in the US than in Costa Rica or Taiwan. This finding may partly reflect higher levels of social stratification in the US than in the other countries, including larger inequities in access to health care and a higher concentration of unhealthy behaviors among individuals occupying lower socioeconomic positions. The variability in associations across countries also underscores the dangers of assuming that these relationships are universal.

Recent studies that incorporate “novel coronary risk factors” suggest that immune and inflammatory markers related to cardiovascular risk may play a more important role than traditional cardiovascular risk factors in mediating the associations between SES and health. For example, many studies in the US and the UK identify substantial SES differences in the levels of such markers as interleukin-6, C-reactive protein, fibrinogen, and tumor necrosis factor-α (Albert, Glynn, Buring, & Ridker 2006; Banks, Marmot, Oldfield, et al. 2006; Friedman & Herd 2010; Gruenewald, Cohen, Matthews, Tracy, & Seeman 2009; Hemingway, Shipley, Mullen, et al. 2003; Koster, Bosma, Penninx, et al. 2006; Marmot, Shipley, Hemingway, et al. 2008; Pollitt, Kaufman, Rose, Diez-Roux, Zeng, & Heiss 2008; Ramsay, Morris, Whincup, et al. 2009; Tabassum, Kumari, Rumley, Lowe, Power, & Strachan 2008). Once again, however, relatively few of these studies assess the impact of these markers on SES differentials in health, and, among those that do, the results vary across studies (due in part to whether behavioral risk factors are included in the model and the specific measure of SES used).

Still, the strong links between these inflammatory and immune markers and various illnesses, chronic conditions, and health-related behaviors, such as smoking and obesity, suggest that their inclusion in biosocial surveys may provide a promising direction for elucidating the pathways linking social disparities to health. In addition, the increasing number of biosocial surveys being fielded in both high and middle-income countries, some of which comprise longitudinal data that permit the identification of incident health conditions and their sequelae, may help researchers to identify true variations in the strength and nature of the relationships between SES and health across different social, economic and cultural settings. At minimum, social scientists need to recognize that the conventional biomarkers of cardiovascular disease used in most studies are only scratching the surface of the complex, multi-system physiological mechanisms through which social disadvantage is likely to get under the skin.

Supplementary Material



This work was supported by the National Institute on Aging (R01AG16790, Goldman and P30AG17265, Crimmins) and the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24HD047879, Goldman). The CRELES survey was supported by the Wellcome Trust (072406/Z/03/Z, Rosero-Bixby). The Health and Retirement Study was sponsored by the National Institute on Aging (U01AG009740, Weir) and was conducted by the University of Michigan. We would like to thank Dana Glei for helpful suggestions on an earlier version of the manuscript.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Contributor Information

Noreen Goldman, Office of Population Research, Princeton University.

Cassio Turra, Departamento de Demografia, Centro de Desenvolvimento e Planejamento Regional de Minas Gerais, Universidade Federal de Minas Gerais.

Luis Rosero-Bixby, Centro Centroamericano de Población, Universidad de Costa Rica.

David Weir, Institute for Social Research, University of Michigan.

Eileen Crimmins, Davis School of Gerontology, University of Southern California.


  • AACE Diabetes Mellitus Clinical Practice Guidelines Task Force. American Association of Clinical Endocrinologists medical guidelines for clinical practice for the management of diabetes mellitus. Endocrine Practice. 2007;13(Suppl 1):3–68. [PubMed]
  • Albert MA, Glynn RJ, Buring J, Ridker PM. Impact of traditional and novel risk factors on the relationship between socioeconomic status and incident cardiovascular events. Circulation. 2006;114(24):2619–2626. [PubMed]
  • Allison DB, Faith MS, Heo M, Kotler DP. Hypothesis concerning the U-shaped relation between body mass index and mortality. American Journal of Epidemiology. 1997;146(4):339–349. [PubMed]
  • Atherton K, Power C. Health inequalities with the National Statistics-Socioeconomic classification: Disease risk factors and health in the 1958 British birth cohort. The European Journal of Public Health. 2007;17(5):486–491. [PubMed]
  • Avendano M, Glymour MM. Stroke disparities in older Americans: Is wealth a more powerful indicator of risk than income and education? Stroke. 2008;39(5):1533–1540. [PMC free article] [PubMed]
  • Banks J, Marmot M, Oldfield Z, Smith JP. Disease and disadvantage in the United States and in England. The Journal of the American Medical Association. 2006;295(17):2037–2045. [PubMed]
  • Barrett-Connor E, Khaw KT, Yen SS. A prospective study of dehydroepiandrosterone sulfate, mortality, and cardiovascular disease. The New England Journal of Medicine. 1986;315(24):1519–1524. [PubMed]
  • Bobak M, Hertzman C, Skodova Z, Marmot M. Socioeconomic status and cardiovascular risk factors in the Czech Republic. International Journal of Epidemiology. 1999;28(1):46–52. [PubMed]
  • Bond Huie SA, Krueger PM, Rogers RG, Hummer RA. Wealth, race, and mortality. Social Science Quarterly (Blackwell Publishing Limited) 2003;84(3):667–684.
  • Brunner EJ, Marmot MG, Nanchahal K, Shipley MJ, Stansfeld SA, Juneja M, et al. Social inequality in coronary risk: Central obesity and the metabolic syndrome. Evidence from the Whitehall II study. Diabetologia. 1997;40:1341–1349. [PubMed]
  • Chang VW, Lauderdale DS. Income disparities in body mass index and obesity in the United States, 1971–2002. Archives of Internal Medicine. 2005;165(18):2122–2128. [PubMed]
  • Dowd JB, Goldman N. Do biomarkers of stress mediate the relation between socioeconomic status and health? Journal of Epidemiology and Community Health. 2006;60(7):633–639. [PMC free article] [PubMed]
  • Elo IT. Social class differentials in health and mortality: Patterns and explanations in comparative perspective. Annual Review of Sociology. 2009;35(1):553–572.
  • Feldman HA, Johannes CB, Araujo AB, Mohr BA, Longcope C, McKinlay JB. Low dehydroepiandrosterone and ischemic heart disease in middle-aged men: Prospective results from the Massachusetts Male Aging Study. American Journal of Epidemiology. 2001;153(1):79–89. [PubMed]
  • Friedman EM, Herd P. Income, education, and inflammation: Differential associations in a national probability sample (the MIDUS study) Psychosomatic Medicine. 2010;72(3):290–300. [PMC free article] [PubMed]
  • Goldman N, Cornman JC, Ming-Cheng Chang. Measuring subjective social status: A case study of older Taiwanese. Journal of Cross-Cultural Gerontology. 2006;21(1):71–89. [PubMed]
  • Goldman N, Turra CM, Glei DA, Lin Y, Weinstein M. Physiological dysregulation and changes in health in an older population. Experimental Gerontology. 2006;41(9):862–870. [PubMed]
  • Gruenewald TL, Cohen S, Matthews KA, Tracy R, Seeman TE. Association of socioeconomic status with inflammation markers in black and white men and women in the Coronary Artery Risk Development in Young Adults (CARDIA) study. Social Science & Medicine. 2009;69(3):451–459. [PMC free article] [PubMed]
  • Harald K, Koskinen S, Jousilahti P, Torppa J, Vartiainen E, Salomaa V. Changes in traditional risk factors no longer explain time trends in cardiovascular mortality and its socioeconomic differences. Journal of Epidemiology and Community Health. 2008;62(3):251–257. [PubMed]
  • Hemingway H, Shipley M, Mullen MJ, Kumari M, Brunner E, Taylor M, et al. Social and psychosocial influences on inflammatory markers and vascular function in civil servants (the Whitehall II study) The American Journal of Cardiology. 2003;92(8):984–987. [PubMed]
  • International Monetary Fund. World economic and financial surveys. 2010. Retrieved July 6, 2010 from http://www.imf.org/external/pubs/ft/weo/2008/02/weodata/index.aspx.
  • Jylhä M, Volpato S, Guralnik JM. Self-rated health showed a graded association with frequently used biomarkers in a large population sample. Journal of Clinical Epidemiology. 2006;59(5):465–471. [PubMed]
  • Kanjilal S, Gregg EW, Cheng YJ, Zhang P, Nelson DE, Mensah G, et al. Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971–2002. Archives of Internal Medicine. 2006;166(21):2348–2355. [PubMed]
  • Kapuku GK, Treiber FA, Davis HC. Relationships among socioeconomic status, stress induced changes in cortisol, and blood pressure in African American males. Annals of Behavioral Medicine. 2002;24(4):320–325. [PubMed]
  • Khang Y, Kim HR. Explaining socioeconomic inequality in mortality among South Koreans: An examination of multiple pathways in a nationally representative longitudinal study. International Journal of Epidemiology. 2005;34(3):630–637. [PubMed]
  • Koster A, Bosma H, Penninx BWJH, Newman AB, Harris TB, van Eijk JTM, et al. Association of inflammatory markers with socioeconomic status. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences. 2006;61(3):284–290. [PubMed]
  • Koster A, Penninx BWJH, Bosma H, Kempen GIJM, Harris TB, Newman AB, et al. Is there a biomedical explanation for socioeconomic differences in incident mobility limitation. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences. 2005;60(8):1022–1027. [PubMed]
  • Kristenson M, Eriksen HR, Sluiter JK, Starke D, Ursin H. Psychobiological mechanisms of socioeconomic differences in health. Social Science & Medicine. 2004;58(8):1511–1522. [PubMed]
  • Love GD, Seeman TE, Weinstein M, Ryff CD. Bioindicators in the MIDUS National Study: Protocol, measures, sample, and comparative context. Journal of Aging and Health. 2010 Pre-publication. [PMC free article] [PubMed]
  • Lu JR, Hsiao WC. Does universal health insurance make health care unaffordable? Lessons from Taiwan. Health Affairs. 2003;22(3):77–88. [PubMed]
  • Lynch JW, Kaplan GA, Cohen RD, Tuomilehto J, Salonen JT. Do cardiovascular risk factors explain the relation between socioeconomic status, risk of all-cause mortality, cardiovascular mortality, and acute myocardial infarction. American Journal of Epidemiology. 1996;144(10):934–942. [PubMed]
  • Marklund N, Peltonen M, Nilsson TK, Olsson T. Low and high circulating cortisol levels predict mortality and cognitive dysfunction early after stroke. Journal of Internal Medicine. 2004;256(1):15–21. [PubMed]
  • Marmot MG, Shipley MJ, Hemingway H, Head J, Brunner EJ. Biological and behavioural explanations of social inequalities in coronary heart disease: The Whitehall II study. Diabetologia. 2008;51:1980–1988. [PMC free article] [PubMed]
  • Martikainen P, Ishizaki M, Marmot MG, Nakagawa H, Kagamimori S. Socioeconomic differences in behavioural and biological risk factors: A comparison of a Japanese and an English cohort of employed men. International Journal of Epidemiology. 2001;30(4):833–838. [PubMed]
  • Mazat L, Lafont S, Berr C, Debuire B, Tessier JF, Dartigues JF, et al. Prospective measurements of dehydroepiandrosterone sulfate in a cohort of elderly subjects: Relationship to gender, subjective health, smoking habits, and 10-year mortality. Proceedings of the National Academy of Sciences of the United States of America. 2001;98(14):8145–8150. [PMC free article] [PubMed]
  • Muennig P, Sohler N, Mahato B. Socioeconomic status as an independent predictor of physiological biomarkers of cardiovascular disease: Evidence from NHANES. Preventive Medicine. 2007;45(1):35–40. [PubMed]
  • National Heart, Lung, and Blood Institute, Obesity Education Initiative. Aim for a healthy weight. information for patients and the public. Retrieved March 17, 2010 from http://www.nhlbi.nih.gov/health/public/heart/obesity/lose_wt/risk.htm.
  • Park HS, Cho S, Song Y, Sung J. Multiple metabolic risk factors and total and cardiovascular mortality in men with low prevalence of obesity. Atherosclerosis. 2006;187(1):123–130. [PubMed]
  • Pollitt RA, Kaufman JS, Rose KM, Diez-Roux AV, Zeng D, Heiss G. Cumulative life course and adult socioeconomic status and markers of inflammation in adulthood. Journal of Epidemiology and Community Health. 2008;62(6):484–491. [PubMed]
  • Population Reference Bureau. World Population Data Sheet. 2009. Retrieved March 23, 2010 from http://www.prb.org/Publications/Datasheets/2009/2009wpds.aspx.
  • Ramsay SE, Morris RW, Whincup PH, Papacosta O, Rumley A, Lennon L, et al. Socioeconomic inequalities in coronary heart disease risk in older age: Contribution of established and novel coronary risk factors. Journal of Thrombosis and Haemostasis. 2009;7(11):1779–1786. [PMC free article] [PubMed]
  • Reddy KK, Rao AR, Reddy TP. Socioeconomic status and the prevalence of coronary heart disease risk factors. Asia Pacific Journal of Clinical Nutrition. 2002;11:98–103. [PubMed]
  • Rehkopf D, Dow W, Rosero Bixby L. Differences in the association of cardiovascular risk factors with education: A comparison of Costa Rica (CRELES) and the United States (NHANES) Journal of Epidemiology and Community Health. 2010;64(9):821–828. [PMC free article] [PubMed]
  • Rosero-Bixby L, Dow WH. Surprising SES gradients in mortality, health, and biomarkers in a Latin American population of adults. The Journals of Gerontology Series B: Psychological Sciences and Social Science 2009 [PMC free article] [PubMed]
  • Seeman TE, Singer BH, Rowe JW, Horwitz RI, McEwen BS. Price of adaptation--allostatic load and its health consequences. MacArthur Studies of Successful Aging. Archives of Internal Medicine. 1997;157(19):2259–2268. [PubMed]
  • Seeman TE, Crimmins E, Huang M, Singer B, Bucur A, Gruenewald T, et al. Cumulative biological risk and socio-economic differences in mortality: MacArthur Studies of Successful Aging. Social Science & Medicine. 2004;58(10):1985–1997. [PubMed]
  • Seidell J, Visscher T. Body weight and weight change and their health implications for the elderly. European Journal of Clinical Nutrition. 2000;54:S33–S39. [PubMed]
  • Seplaki CL, Goldman N, Glei D, Weinstein M. A comparative analysis of measurement approaches for physiological dysregulation in an older population. Experimental Gerontology. 2005;40(5):438–449. [PubMed]
  • Siegrist J, Marmot M. Health inequalities and the psychosocial environment—two scientific challenges. Social Science & Medicine. 2004;58(8):1463–1473. [PubMed]
  • Smith KV, Goldman N. Socioeconomic differences in health among older adults in Mexico. Social Science & Medicine. 2007;65(7):1372–1385. [PubMed]
  • StataCorp. Stata User’s Guide: Release 10.0. Stata Corporation; Colllege Station, TX: 2008.
  • Steptoe A, Kunz-Ebrecht S, Owen N, Feldman PJ, Willemsen G, Kirschbaum C, et al. Socioeconomic status and stress-related biological responses over the working day. Psychosomatic Medicine. 2003;65(3):461–470. [PubMed]
  • Steptoe A, Marmot M. The role of psychobiological pathways in socioeconomic inequalities in cardiovascular disease risk. European Heart Journal. 2002;23(1):13–25. [PubMed]
  • Tabassum F, Kumari M, Rumley A, Lowe G, Power C, Strachan DP. Effects of socioeconomic position on inflammatory and hemostatic markers: A life-course analysis in the 1958 British birth cohort. American Journal of Epidemiology. 2008;167(11):1332–1341. [PubMed]
  • Third Report of the National Cholesterol Education Program (NCEP) Expert Panel. Detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III) Final Report No. 02-5215. National Institutes of Health, National Heart, Lung, and Blood Institute; 2002. [PubMed]
  • UNESCO Institute for Statistics. Education in Costa Rica. 2009. UIS statistics in brief.
  • Unger J, De Paepe P, Buitron R, Soors W. Costa Rica: Achievements of a heterodox health policy. American Journal of Public Health. 2008;98(4):636–643. [PMC free article] [PubMed]
  • United Nations Educational, Scientific and Cultural Organization. Appendix 2 School Life Expectancy for Age 6 and Above by Gender.
  • Weinstein M, Vaupel JW, Wachter KW, editors. Biosocial surveys: Committee on advances in collecting and utilizing biological indicators and genetic information in social science surveys. Washington, D.C: The National Academies Press; 2008. [PubMed]
  • Weir D. Elastic powers: The integration of biomarkers into the health and retirement study. In: Weinstein M, Vaupel JW, Wachter KW, editors. Biosocial surveys. Washington, D.C: The National Academies Press; 2008. pp. 78–95.
  • Whitworth JA. World Health Organization (WHO) International Society of Hypertension (ISH) statement on management of hypertension. Journal of Hypertension. 2003;21(11):1983–1992. [PubMed]
  • World Health Organization (WHO); WHO. Obesity: Prevent and managing the global epidemic. Geneva: WHO; 1998.
  • Song Yun-Mi, Ferrer RL, Cho Sung-il, Sung J, Ebrahim S, Smith GD. Socioeconomic status and cardiovascular disease among men: The Korean National Health Service Prospective Cohort Study. American Journal of Public Health. 2006;96(1):152–159. [PMC free article] [PubMed]
  • Zajacova A, Goldman N, Rodriguez G. Unobserved heterogeneity can confound the effect of education on mortality. Mathematical Population Studies. 2009;16(2):153–173. [PMC free article] [PubMed]
  • Zhang Q, Wang Y. Socioeconomic inequality of obesity in the United States: Do gender, age, and ethnicity matter? Social Science & Medicine. 2004;58(6):1171–1180. [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Cited in Books
    Cited in Books
    PubMed Central articles cited in books
  • Compound
    PubChem Compound links
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...