Exceptions are interesting: They make us ask “why?” Why is this case different? From a scientific perspective—at least in principle—we start with a theory or empirical generalization, and seek to reject it; too often, however, as researchers we seek evidence to confirm our hypotheses. It’s only when we find enough exceptions that we feel compelled to reject the rule or significantly amend it. Kahneman (2011, p. 81), as always, makes the point regarding “a deliberate search for confirming evidence, known as positive test strategy,” with elegance and parsimony: “Contrary to the rules of philosophers of science, who advise testing hypotheses by trying to disprove them, people (and scientists, quite often) seek data that are likely to be compatible with the beliefs they currently hold.”
Here we sketch out a case for reconsidering the theoretical motivation for much of the recent biosocial survey efforts. We concentrate on just a few points. First, we present evidence that suggests that health disparities by socioeconomic status (SES) are not without exceptions. Second, we argue that we have only weak evidence showing that biomarkers—at least the ones that are most commonly collected in population-level biosocial studies—mediate the relationship between social status and health. Finally, we focus on the allostatic load paradigm. Allostatic load has been an important guiding framework for much of the biosocial research efforts; correspondingly, it has been widely critiqued. Because the criticisms are well rehearsed—although often ignored in practice—we only briefly discuss some of the vulnerabilities with its application “on the ground.”
These considerations lead us to argue that it is time for our biodemographic investigations to incorporate and test the kinds of theoretical
scaffolding that sociology and evolutionary biology can provide. We suggest a few directions where we believe such theoretical articulation might be productive, but especially in light of our own deficiencies in those areas, we encourage the biodemographic community to be more active in reaching out to the students of those disciplines. Gruenewald’s chapter provides a well-organized overview of much of the field, highlighting some of the exciting potential for future work. Our excitement regarding that potential is perhaps more tempered by some of our experience.
THE SES-HEALTH GRADIENT DOES NOT APPEAR TO BE UNIVERSAL
As noted by Gruenewald, much of the literature documenting links between SES and health (or mortality) suggests that the lower the position in a social hierarchy—typically measured as education, income, or occupational status—the higher the risk of poorer health or earlier death. Much of the evidence for this generalization comes from higher income countries—and even among higher income countries there are inconsistent results—but an increasing body of evidence from middle-income or more recently developed countries appears to show less consistent or weaker patterns (Goldman et al., 2011). Some of these “exceptions” have been documented only recently (Rosero-Bixby and Dow, 2009; Smith and Goldman, 2007), but others, including higher income countries, have been noted for some time. Sweden is such a case. Vagero and Landberg (1989, cited in Wilkinson, 1996, Figure 5.8), for example, compared age-standardized death rates across social classes in Sweden with England and Wales, and Leon et al. (1992; cited in Wilkinson, 1996) did a similar analysis of infant mortality. Both pictures show the expected gradient in England and Wales; the picture in Sweden, however, shows little evidence of an SES gradient. Japanese data for the 1990s (Hirokawa, Tsutusmi, and Kayaba, 2006) show no effect of education (assessed as age at completion of education) or employment status on all-cause mortality for persons aged 60 and older, and no effect of employment status on all-cause mortality of persons aged 59 and younger. Age at completion of education had a discernible effect on mortality for women below age 59 who finished school before age 15.
Age may be one important factor. There is ongoing debate regarding whether the effects of SES on health are likely to increase with age—as the advantages of higher status may accumulate throughout the life cycle—or whether they are expected to decline—as biological frailty dominates social influences and medical care becomes more widely accessible, for example, through Medicare for persons aged 65 and older in the United States (Dannefer, 2003; House et al., 1994). Most empirical work supports the second argument: SES disparities in health generally decline from
middle to older ages (Smith and Goldman, 2007; Zajacova, Goldman, and Rodriguez, 2009). Another age-related effect may also be operating: generally, we would expect the least healthy to die first so one might expect less variability in health as age increases. Either way, surveys restricted to older adults are less likely than those based on a broader and younger age range to identify statistically significant (or meaningful) social disparities in health outcomes.
Another—albeit hotly contested—explanation for these apparent anomalies may be related to the extent of social inequality. In the aggregate, some work has shown that income inequality appears to be (inversely) associated with life expectancy in wealthy countries (de Vogli et al., 2005) and, in the United States, directly related to mortality rates in metropolitan areas (Ash and Robinson, 2009; Lynch et al., 1998). Other analyses, in some instances by the same authors, suggest that the apparent inverse relationship disappears when additional underlying factors are taken into account (Deaton and Lubotsky, 2009; Deaton and Paxson, 2004; Lynch et al., 2000, 2004a,b).
All of this is not to say that SES-associated gradients do not exist. Clearly they do, in some places, at some times, perhaps even in most places—at least in modern times. Any generalization is almost certain to have exceptions, and a very extensive literature would essentially guarantee that we could find at least some support for any claim or counterclaim. Should we focus simply on survival or some measure(s) of health? How—and when—should we measure the outcomes? SES may have a very different relationship with survival (or health) at younger ages from the relationships it has at older ages; some dimensions of health may be more highly associated than others. “When” may also matter not only in terms of age, but also in terms of historical time or situation in the epidemiological transition. For example, historical data for the United States reveal few social class differences in child mortality in the late 19th century (e.g., the children of physicians had death rates close to the national average), but, as health beliefs changed and knowledge of hygienic practices spread during subsequent decades, socioeconomic gradients in mortality widened (Preston and Haines, 1991). And of course, we could always find differences among studies in the ways SES is realized or how health is measured.
Still, we would hope that a “good” generalization would be robust. Len Syme says it well (personal communication, 2012):
We are rightly concerned about defining and measuring variables precisely and perfectly but, to me, the most important variables withstand our imprecision and vagueness. For example, if we used different definitions and different methods in different populations and still always find the same results, I think we have a really important variable. Social class provides a good example of this. We don’t really know what social class is and we
measure it in many different ways. But that doesn’t seem to matter; Social class turns out to be a very important variable because it is always related to the outcomes we study. Social support provides another example. These variables are so important that they withstand our clumsy attempts at definition and measurement.
We almost agree: Our point is that the exceptions to the “always” are interesting because they suggest we may be missing something that should motivate a search for richer more nuanced explanations of those findings. The exceptions remind us that we shouldn’t expect to see the same relationships everywhere.
WEAK EVIDENCE THAT BIOMARKERS MEDIATE THE SES–HEALTH LINK
Here we raise two concerns. First, it is not entirely clear whether an unambiguous relationship between social conditions—broadly construed as position in social hierarchies, social relationships, and networks—and biomarkers has been documented. And second, the extent to which currently obtainable biomarkers mediate the relationship between social conditions and health appears to be an open question. We are not suggesting some magical connection. At some level, we are biological reductionists: We accept that the association must be mediated through physiological pathways, and we caution only that a convincing case has not yet been made. Gruenewald points to the desire to understand how social conditions “get under the skin” as a motivation for the addition of biomarker collection to epidemiological and social and demographic studies. Although she identifies some of the apparent vulnerabilities of these efforts, our overall impression from her commentary is that she is quite optimistic. For a variety of reasons, we have a more guarded view of the landscape. We have concerns about choice of biomarkers, dealing with complex interactions between genetic endowment and environment, the large numbers of pathways for which we would want biomarkers, and finally, we suspect that the physiological influences linking SES to health are comprised of huge numbers of potentially interactive effects, most of which are not observable or measurable. Other measurement issues range from determining how, when—or how often—to measure biomarkers to how to measure environmental influences, especially past environmental influences.
Evidence linking social conditions and biomarkers is not unambiguous. As Gruenewald discusses, even a relatively well-measured marker, such as blood pressure with established ties to disease processes, exhibits markedly different relationships with SES across studies or even within studies depending on sex or measure of SES (Goldman et al., 2011). Other
biomarkers raise even more complex issues of assay comparability and, more fundamentally, the processes that the markers reflect, processes that may differ across time and setting.
Work on the relationship between social factors and markers of immune function or inflammation serves as a good example of such problems. Recent reviews of the literature by Uchino (2006) and Kiecolt-Glaser, Gouin, and Hantsoo (2010) provide insights into factors that make it difficult to generalize about the links. One example is the fact that social interactions typically entail both positive and negative aspects; another is that commonly measured biomarkers of inflammation (IL-6, for example, which can have both pro- and anti-inflammatory influences [Uchino, 2006]) have highly complex mechanisms that could easily be misinterpreted in observational study designs.
One account, proposed by Hillard Kaplan during the course of the workshop, suggests that some of the difficulties could lie in the causal pathways that underlie the inflammatory markers (Kaplan, personal communication, 2011). For example, the history of exposure to infection might influence inflammation throughout the life course. More generally, Kaplan suggested that the causal pathways would be likely to differ not only across environmental conditions, but also with age so that researchers might confound adaptive aspects of aging with potentially correlated, but not necessarily causal, social exposures. Thus, for instance, higher levels of inflammation might be an adaptive response to age-related changes rather than a marker of poor regulation.
Our own recent work in collaboration with Carol Ryff and Yu-Hsuan Lin (Glei et al., 2012a) using U.S. and Taiwanese data provides little encouragement. We examined the relation between two components of social relationships—perceived support and social integration—and six inflammatory markers. Results yielded only weak evidence of a link between the biomarkers and the social relationships. Along the lines of Kaplan’s suggestion, one might expect that exposure to infection, especially when today’s older adults were children, would be a more important promoter of inflammation in Taiwan than in the United States. If so, that could weaken the potential effect of social relationships in Taiwan. However, even this very plausible suggestion is not supported by the data in this instance: If anything, the association between social relationships and inflammation appeared stronger in Taiwan than in the United States. We recognize that our data cannot support a conclusive test: We do not have direct information on childhood exposure to infectious disease in Taiwan, although, perhaps, a comparison of inflammation in cohorts who were born early versus late in the epidemiological transition might shed some light. We also do not know whether, as discussed earlier, exposure would have pro- or anti-inflammatory effects in adulthood. To move forward, we need to have
hypotheses that direct our attention to the complex interactions and links among social organization, physical conditions, macro-level change in these factors, and individual-level response to exposures. Such hypotheses present heavy—perhaps insupportable—demands on any data collection initiative and will almost certainly require an approach that articulates data across time and place. Such worthwhile efforts at integration, as we discuss later, face their own challenges.
We are also collaborating with colleagues using data from the Survey on Stress, Aging and Health in Russia, a survey of Muscovites aged 55 and older (Glei et al., 2012b; Shkolnikova et al., 2009). Russia might be the “poster child case” for establishing an association between social disparities and mortality: On the one hand, the most greatly disadvantaged Russians (especially men) suffered the greatest declines in life expectancy during the mortality crisis; on the other, highly educated Russians experienced an increase in life expectancy (Murphy et al., 2006; Shkolnikov et al., 2006). If a link between SES and biomarkers corresponding to a link between SES and mortality could be documented anywhere, we expected to find strong evidence in Russia. Indeed, we found substantial educational disparities in physiological dysregulation based on 20 biomarkers. However, more detailed analysis revealed that the size of the differentials varied across systems. Both sexes exhibited a large educational disparity in standard cardiovascular and metabolic factors, but heart rate parameters (based on 24h ECG) and inflammation showed substantial differences only in men. These results are consistent with the excess cardiovascular mortality that is a major contributor to high levels of mortality among Russians, particularly men. Yet, social disparities in neuroendocrine dysregulation were negligible in both sexes. If social disparities in allostatic load and in health outcomes reflect a differential burden of stress, it seems surprising to find so little social variation in these stress hormones, although we recognize the many measurement issues surrounding the collection of these markers.
Whether the current battery of biomarkers actually mediates the relation between social conditions and health is an even more vexed question. Unlike several other studies (see next sentence), the Russia data do show that the biomarkers—accumulated across systems—explain a substantial proportion (albeit only about one-third at best) of the variation in health across SES groups. Other studies that have examined whether biomarkers mediate social disparities in self-assessed health status or physical functioning show that the biomarkers explain only a small proportion of the socioeconomic differentials in Taiwan (Dowd and Goldman, 2006; Goldman et al., 2011; Hu et al., 2007), relatively little of the variation in the United States (Goldman et al., 2011; Koster et al., 2005), and none of the SES gap in Costa Rica (Goldman et al., 2011). As Gruenewald notes in her chapter, “evidence demonstrating that social disparities in biomarkers underlie
social disparities in actual health outcomes” is needed. To date, such evidence is sparse at best, and must be stacked up against a growing body of null or inconsistent findings—at least from population-based studies. The likelihood that null findings are underrepresented because of publication bias simply serves to underscore this point. Overall then, we have not done well at explaining the physiological pathways linking SES to health.
ALLOSTATIC LOAD—A FEW CONCERNS
Never underestimate the power of a narrative (Kahneman, 2011, p. 81)—and the story behind allostatic load is compelling. A recent review by Juster, McEwen, and Lupien (2010, p. 3) provides a simple summary of the plot: “Allostatic load (AL) represents the ‘wear and tear’ the body experiences when repeated allostatic responses are activated during stressful situations” (McEwen and Steller, 1993). In turn, allostatic response (Juster, McEwen, and Lupien, 2010, p. 2) is the “process whereby an organism maintains physiological stability by changing parameters of its internal milieu by matching them appropriately to environmental demands” (Sterling and Eyer, 1988). The 2010 Juster et al. review summarizes some 58 studies of AL; we estimate that—perhaps—15 of them include measures of stressors or perceived stress. Five of those were based on the Taiwan Social Environment and Biomarkers of Aging Study data and yielded only modest support for links between AL and stress (Gersten, 2008; Glei et al., 2007; Seeman et al., 2004; Weinstein et al., 2003). Studies in the United States have also yielded some evidence of a modest association (Roepke et al., 2011; von Känel et al., 2003) but others have found mixed results (Gallo et al., 2011; Mair, Cutchin, and Kirsten Peek, 2011) or no association (Seeman et al., 2002). Still other research found a weak relationship in Australia (Clark, Bond, and Hecker, 2007) and Germany (Schnorpfeil et al., 2003), and a modest association in China (Sun et al., 2007) and Sweden (Gustafsson et al., 2011). In short, while there is substantial evidence that multisystem dysregulation—to use that term rather than AL, which implies a link to stressful experience—is related to many health outcomes, its links to stressful experience are not well established.
Concerns with allostatic load are nothing new. The National Institute on Aging (NIA) Exploratory Workshop on Allostatic Load, held under the auspices of the Behavioral and Social Research Program, NIA, was convened November 29-30, 2007, in part to shed some light on the issues. Background materials from the workshop provide a laundry list of such concerns (Nielson, Seeman, and Hahn, 2007). A full discussion of the concerns is not what we would want to accomplish here, but we note that the participants raised issues with (among others): how “stress” is defined (Cacioppo, Crimmins, Epel, Goldman); the choice of biomarkers
that capture dynamics or reflect cumulative dysregulation (Cohen, Coles, Epel, Goldman); and understanding the role of the timing of exposure (Maestripieri)—a question also raised by Gruenewald. One might add assay comparability across time and place, the need for a developmental approach that incorporates exposure and health across the life course, and the various logistical and financial hurdles involved in incorporating welldesigned biomarker collection in population surveys.
The definition and measurement of “stress” is a particularly thorny problem (Cohen, Kessler, and Underwood, 1997; Monroe, 2008). A related issue pertains to how “stress” fits into the allostatic load framework. One viewpoint suggests that allostatic load provides a measure of physiological stress. This perspective—which tautologically links stress to allostatic load—fails to provide us with testable hypotheses regarding the impact of life challenges or other environmental factors on physiological dysregulation. An alternative framework, which underlies much of the research described in this paper, posits that dysregulation is a result of prolonged or repeated exposure to life stressors. In this case, there is a testable relationship, but one that has not been studied systematically and, to date, has yielded only weak evidence of causal linkages.
WHERE CAN WE GO FROM HERE?
We are not yet ready to deny more generally the utility of documenting physiological parameters of a population, but we would argue that future forays into biosocial survey data collection need to be grounded in well-formulated theory. No one is advocating throwing the wheat away with the chaff, and we recognize that it may be too early to decide what to keep and what to toss. We see several areas for development. Gruenewald talks about geographic variation and we agree that it is a potentially fruitful area for investigation. “Geographic” encompasses a multitude of possible explanatory factors including variation in environmental conditions, gender roles and relations, epidemiologic history, social structures and institutions, culture, developmental histories, and genetic endowment. We have now amassed an impressive array of biosocial studies across a wide range of geographies, and now the question is how can we best exploit these data.
If we want to move beyond purely descriptive “comparisons” to understand the deeper, possibly causal explanations of variation—and it does seem like a worthy goal—we will need better theory to inform our investigations. It is here that we see a large potential contribution from sociology, from both cultural and biological anthropology, from psychology, and from evolutionary biology. As a field, we are spending too much time talking among ourselves. This commentary has noted or implied a few areas from which to begin: we need help understanding, for example, how
social stratification varies across geographies, how social institutions and structures mutually contribute to and reinforce each other’s formation and perpetuation (Sewell, 1992), and the role of priming in relative deprivation (Kahneman, 2011). We have data that will allow us to perform similar analyses in Costa Rica, Russia, Taiwan, and the United States, but how do we explain differences when we find them? As noted by Goldman and her colleagues (2011, p. 313):
Despite a justified appeal for international comparisons of social gradients in health that integrate biological mechanisms, 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 differences in sample size.
These questions are not only limited to different geographies, but also apply to group differences more generally. How can we explain different relationships among variables when we find them between, for example, men and women? Underlying physiological differences by sex may be only one factor. As Gruenewald says, social factors interact with biology in complex ways: those differences between men and women are almost certain to also have a basis in the social interpretation and expectations for each sex. Similarly, we would look for deeper explanations of black/white differentials.
We have also mentioned the need for both epidemiologic history and evolutionary biology in our discussion of inflammation, but one could easily imagine that those questions factor in to just about any physiological– social link that we would want to examine. Overall, we would advocate for comparative studies that bring together diverse explanations for the links between physiology and social conditions. Are there ways to test (i.e., reject) the theories? If—as seems likely—additional data collection initiatives continue to be funded, we would advocate for carefully targeted, theoretically driven studies.
So, is the glass half empty or is it half full? As always, the answer is “both.” We remain both skeptical and cautiously optimistic.
Ash, M., and Robinson, D.E. (2009). Inequality, race, and mortality in U.S. cities: A political and econometric review of Deaton and Lubotsky. Social Science and Medicine, 68(11), 1,909-1,913.
Clark, M.S., Bond, M.J., and Hecker, J.R. (2007). Environmental stress, psychological stress, and allostatic load. Psychology, Health and Medicine, 12(1), 18-30.
Cohen, S., Kessler, R.C., and Underwood, L. (1997). Measuring Stress: A Guide for Health and Social Scientists. New York: Oxford University Press.
Dannefer, D. (2003). Cumulative advantage/disadvantage and the life course: Cross-fertilizing age and social science theory. The Journals of Gerontology: Series B Psychological Sciences and Social Sciences, 58(6), S327-S337.
De Vogli, R., Mistry, R., Gnesotto, R., and Cornia, G.A. (2005). Has the relation between income inequality and life expectancy disappeared? Evidence from Italy and top industrialized countries. Journal of Epidemiology and Community Health, 59(2), 158-162.
Deaton, A., and Lubotsky, D. (2009). Income inequality and mortality in U.S. cities: Weighing the evidence. A response to Ash. Social Science and Medicine, 68(11), 1,914-1,917.
Deaton, A., and Paxson, C. (2004). Mortality, income, and income inequality over time in Britain and the United States. In D. Wise (Ed.), Perspectives on the Economics of Aging. Chicago: University of Chicago Press.
Dowd, J.B., and Goldman, N. (2006). Do biomarkers of stress mediate the relation between socioeconomic status and health? Journal of Epidemiology and Community Health, 60(7), 633-639.
Gallo, L.C., Jiménez, J.A., Shivpuri, S., Espinosa De Los Monteros, K., and Mills, P.J. (2011). Domains of chronic stress, lifestyle factors, and allostatic load in middle-aged Mexican-American women. Annals of Behavioral Medicine, 41(1), 21-31.
Gersten, O. (2008). Neuroendocrine biomarkers, social relations, and the cumulative costs of stress in Taiwan. Social Science & Medicine, 66(3), 507-519.
Glei, D.A., Goldman, N., Chuang, Y.L., and Weinstein, M. (2007). Do chronic stressors lead to physiological dysregulation? Testing the theory of allostatic load. Psychosomatic Medicine, 69(8), 769-776.
Glei, D.A., Goldman, D.P., Ryff, C., Lin, Y.-H., and Weinstein, M. (2012a). Social relationships and inflammatory markers: A comparative analysis of Taiwan and the U.S. Social Science & Medicine, 74(12), 1,891-1,899.
Glei, D.A., Goldman, N., Shkolnikov, V.M., Jdanov, D., Shalnova, S., Shkolnikova, M., Vaupel, J.W., and Weinstein, M. (2012b). To What Extent Do Biological Markers Account for the Large Social Disparities in Health in Moscow? Paper presented at the Population Association of America. San Francisco, May 3-5. Available: http://paa2012.princeton.edu/download.aspx?submissionId=120358 [September 13, 2012].
Goldman, N., Turra, C.M., Rosero-Bixby, L., Weir, D., and Crimmins, E. (2011). Do biological measures mediate the relationship between education and health: A comparative study. Social Science and Medicine, 72(2), 307-315.
Gustafsson, P.E., Janlert, U., Theorell, T., Westerlund, H., and Hammarström, A. (2011). Social and material adversity from adolescence to adulthood and allostatic load in middleaged women and men: Results from the northern Swedish cohort. Annals of Behavioral Medicine, 1-12.
Hirokawa, K., Tsutusmi, A., and Kayaba, K. (2006). Impacts of educational level and employment status on mortality for Japanese women and men: The Jichi Medical School Cohort Study. European Journal of Epidemiology, 21(9), 641-651.
House, J.S., Lepkowski, J.M., Kinney, A.M., Mero, R.P., Kessler, R.C., and Herzog, A.R. (1994). The social-stratification of aging and health. Journal of Health and Social Behavior, 35(3), 213-234.
Hu, P., Wagle, N., Goldman, N., Weinstein, M., and Seeman, T.E. (2007). The associations between socioeconomic status, allostatic load and measures of health in older Taiwanese persons: Taiwan Social Environment and Biomarkers of Aging Study. Journal of Biosocial Science, 39(4), 545-556.
Juster, R.P., McEwen, B.S., and Lupien, S.J. (2010). Allostatic load biomarkers of chronic stress and impact on health and cognition. Neuroscience and Biobehavioral Reviews, 35(1), 2-16.
Kahneman, D. (2011). Thinking, Fast and Slow (1st ed.). New York: Farrar, Straus and Giroux.
Kiecolt-Glaser, J.K., Gouin, J.P., and Hantsoo, L. (2010). Close relationships, inflammation, and health. Neuroscience and Biobehavioral Reviews, 35(1), 33-38.
Koster, A., Penninx, B.W., Bosma, H., Kempen, G.I., Harris, T.B., Newman, A.B., Rooks, R.N., Rubin, S.M., Simonsick, E.M., van Eijk, J.T., and Kritchevsky, S.B. (2005). Is there a biomedical explanation for socioeconomic differences in incident mobility limitation? The Journals of Gerontology: Series A Biological Sciences and Medical Sciences, 60(8), 1,022-1,027.
Lynch, J.W., Kaplan, G.A., Pamuk, E.R., Cohen, R.D., Heck, K.E., Balfour, J.L., and Yen, I.H. (1998). Income inequality and mortality in metropolitan areas of the United States. American Journal of Public Health, 88(7), 1,074-1,080.
Lynch, J.W., Smith, G.D., Kaplan, G.A., and House, J.S. (2000). Income inequality and mortality: Importance to health of individual income, psychosocial environment, or material conditions. British Medical Journal, 320(7,243), 1,200-1,204.
Lynch, J., Davey Smith, G., Harper, S., and Hillemeier, M. (2004a). Is income inequality a determinant of population health? Part 2. U.S. national and regional trends in income inequality and age- and cause-specific mortality. The Milbank Quarterly, 82(2), 355-400.
Lynch, J., Smith, G.D., Harper, S., Hillemeier, M., Ross, N., Kaplan, G.A., and Wolfson, M. (2004b). Is income inequality a determinant of population health? Part 1. A systematic review. Milbank Quarterly, 82(1), 5-99.
Mair, C.A., Cutchin, M.P., and Kristen Peek, M. (2011). Allostatic load in an environmental riskscape: The role of stressors and gender. Health and Place, 17(4), 978-987.
McEwen, B.S., and Stellar, E. (1993). Stress and the individual: Mechanisms leading to disease. Archives of Internal Medicine, 153(18), 2,093-2,101.
Monroe, S.M. (2008). Modern approaches to conceptualizing and measuring human life stress. Annual Review of Clinical Psychology, 4, 33-52.
Murphy, M., Bobak, M., Nicholson, A., Rose, R., and Marmot, M. (2006). The widening gap in mortality by educational level in the Russian Federation, 1980-2001. American Journal of Public Health, 96(7), 1,293-1,299.
Preston, S.H.H., and Haines, M. (1991). Fatal Years: Child Mortality in Late Nineteenth-Century America. Princeton, NJ: Princeton University Press.
Roepke, S.K., Mausbach, B.T., Patterson, T.L., Von Känel, R., Ancoli-Israel, S., Harmell, A.L., Dimsdale, J.E., Aschbacher, K., Mills, P.J., Ziegler, M.G., Allison, M., and Grant, I. (2011). Effects of Alzheimer caregiving on allostatic load. Journal of Health Psychology, 16(1), 58-69.
Rosero-Bixby, L., and Dow, W.H. (2009). 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 Sciences, 64(1), 105-117.
Schnorpfeil, P., Noll, A., Schulze, R., Ehlert, U., Frey, K., and Fischer, J.E. (2003). Allostatic load and work conditions. Social Science and Medicine, 57(4), 647-656.
Seeman, T.E., Singer, B.H., Ryff, C.D., Dienberg Love, G., and Levy-Storms, L. (2002). Social relationships, gender, and allostatic load across two age cohorts. Psychosomatic Medicine, 64(3), 395-406.
Seeman, T., Glei, D., Goldman, N., Weinstein, M., Singer, B., and Lin, Y.H. (2004). Social relationships and allostatic load in Taiwanese elderly and near elderly. Social Science and Medicine, 59(11), 2,245-2,257.
Sewell, W.H. (1992). A theory of structure: Duality, agency, and transformation. American Journal of Sociology, 98(1), 1-29.
Shkolnikov, V.M., Andreev, E.M., Jasilionis, D., Leinsalu, M., Antonova, O.I., and McKee, M. (2006). The changing relation between education and life expectancy in central and eastern Europe in the 1990s. Journal of Epidemiology and Community Health, 60(10), 875-881.
Shkolnikova, M., Shalnova, S., Shkolnikov, V.M., Metelskaya, V., Deev, A., Andreev, E., Jdanov, D., and Vaupel, J.W. (2009). Biological mechanisms of disease and death in Moscow: Rationale and design of the Survey on Stress Aging and Health in Russia (SAHR). BMC Public Health, 9, 293.
Smith, G.D., and Egger, M. (1996). Commentary: Understanding it all—health, meta-theories, and mortality trends. BMJ, 313(7,072), 1,584-1,585.
Smith, K.V., and Goldman, N. (2007). Socioeconomic differences in health among older adults in Mexico. Social Science and Medicine, 65(7), 1,372-1,385.
Sterling, P., and Eyer, J. (1988). Allostasis: A new paradigm to explain arousal pathology. In S. Fisher and J.T. Reason (Eds.), Handbook of Life Stress, Cognition, and Health (pp. 629-649). New York: Wiley.
Sun, J., Wang, S., Zhang, J.Q., and Li, W. (2007). Assessing the cumulative effects of stress: The association between job stress and allostatic load in a large sample of Chinese employees. Work and Stress, 21(4), 333-347.
Uchino, B. (2006). Social support and health: A review of physiological processes potentially underlying links to disease outcomes. Journal of Behavioral Medicine, 29(4), 377-387.
Von Känel, R., Dimsdale, J.E., Patterson, T.L., and Grant, I. (2003). Acute procoagulant stress response as a dynamic measure of allostatic load in Alzheimer caregivers. Annals of Behavioral Medicine, 26(1), 42-48.
Weinstein, M., Goldman, N., Hedley, A., Yu-Hsuan, L., and Seeman, T. (2003). Social linkages to biological markers of health among the elderly. Journal of Biosocial Science, 35(3), 433-453.
Wilkinson, R.G. (1996). Unhealthy Societies: The Afflictions of Inequality. New York: Routledge.
Zajacova, A., Goldman, N., and Rodriguez, G. (2009). Unobserved heterogeneity can confound the effect of education on mortality. Mathematical Population Studies, 16(2), 153-173.