MacArthur SES & Health Network
MacArthur SES & Health Network

printable version

Behavioral and Social Research Program

National Institute on Aging
National Institutes of Health

NIA Exploratory Workshop on Allostatic Load

Washington, DC, November 29-30, 2007

Background Materials and Statements from November 2007 Workshop Participants

Compiled November 19, 2007

Lis Nielsen, Ph.D.
Psychological Development and Integrative Science
Telephone: 301-402-4156

Teresa Seeman, Ph.D (UCLA)
Consultant to BSR

Anneliese Hahn, M.S.
Research Program Analyst
Telephone: 301-402-4447

Background Statement for NIA Exploratory Workshop on Allostatic Load
Lis Nielsen, NIA/BSR
Teresa Seeman, UCLA

Chapter Contents

  1. Introduction
  2. Workshop Goals
  3. Background
    Aging and allostatic load
    Mechanisms and measurement
    State of the evidence
  4. Topics to be addressed in the workshop and papers
    Conceptual and theoretical issues
    Methodological: measurement of allostatic load
    Additional measurement issues include:
  5. References


The concept of allostatic load has served as a framework for a large body of research on the integrative health psychology, epidemiology, and demography of aging. It is based on the hypothesis that there is a cumulative physiological risk associated with exposure to psychosocial stressors over the life-course. Among the attractions of such a concept is the existing body of evidence indicating that many psychosocial stressors appear to have small to modest associations with multiple different biological risk factors, reflecting links to most of the known major regulatory systems (e.g., cardiovascular, immune, HPA, SNS) Initial empirical work based on various cumulative indices of physiological risk has provided evidence consistent with the idea that greater cumulative dysregulation is associated with significantly greater risks for subsequent disease (cardiovascular disease), declines in physical and cognitive functioning and overall mortality (Seeman et al, 1997; Seeman et al, 2001; Karlamangla et al, 2002; Geronimus et al, 2006). Research has also documented that psychosocial conditions previously associated with greater morbidity and mortality (e.g., lower socio-economic status and poorer social engagement) are also associated with greater cumulative burdens of physiological dysregulation in multiple systems (Seeman et al, 2002; Seeman et al, 2004 Kubzansky et al 1999; Hu et al, 2006). Cumulative indices of allostatic load have also been positively related to measures of psychosocial stress in young adolescents (Evans et al, 2007) as well as symptoms of post-traumatic stress disorder (PTSD) in mothers of pediatric cancer survivors (Glover et al, 2006) and adverse perinatal outcomes (Shannon et al, 2007).

back to top

Workshop Goals

Important questions remain, however, regarding the conceptualization and operationalization of allostatic load as well as the overall utility of the concept in efforts to better understand trajectories of health and aging. The Behavioral and Social Research Program (BSR) at the National Institute on Aging seeks to advance research on biopsychosocial pathways of resilience and vulnerability to late life disease through a workshop on conceptual and methodological issues surrounding the concepts of allostatic load and cumulative physiological risk more generally. The goal of the 2007 workshop is to bring together scientists from diverse disciplines who share an interest in understanding stress-health relationships from a life-course perspective, but who may differ in their approaches and commitment to the allostatic load model. We conceive of this group as a collaborative team whose interests are focused on what is needed to advance behavioral and social research on aging within this topic area.

The output of this workshop should include widely distributed publications that will serve as references for established and new researchers in the fields of behavioral and social aging research along with recommendations regarding needed research to advance our understanding of the biological pathways through which our life experiences impact on health and aging.

back to top


Aging and allostatic load

Aging is associated with increased risk for most forms of disability and chronic disease, with the majority of older adults experiencing two or more chronic conditions by the time they reach old age (Singer, Ryff, & Seeman, 2004). It has been hypothesized that cumulative lifetime exposure to social, psychological or environmental stressors increases the risk of multiple age-related health problems by disrupting the physiological regulatory systems that mediate the stress response. Allostatic load has been put forth as a model for how features of the psychosocial environment “get under the skin” and give rise to disease. The model (based on the concept of biological adaptation to duress first proposed by Cannon (1932) and Selye (1956, 1974), developed by McEwen and colleagues (McEwen and Stellar, 1993; McEwen, 1998; McEwen and Seeman, 1999; and elaborated most recently in a volume edited by Jay Schulkin, 2004) proposes that a key mediator of increasing risk for disease is the dysregulation of systems designed to balance the organism’s responses to environmental demands. Exposure to stress elicits adaptive physiological responses in regulatory systems including the hypothalamic pituitary axis (HPA), the sympathetic (SNS) and parasympathetic (PNS) branches of the autonomic nervous system, and the cardiovascular and immune systems. Allostasis (related to homeostasis) is the adaptive maintenance of vitality in these systems via neuromodulation of motive states and behaviors in response to changing environmental circumstances. Allostatic load refers to the cumulative biological wear and tear that can result from excessive cycles of response (i.e., too frequent and/or of inappropriate duration or scope) in these systems as they seek to maintain allostasis in the face of environmental challenge. According to the theory, as these systems become taxed and dysregulated, they begin to exhibit imbalances in the primary neural mediators of the stress response, such as glucocorticoids, catecholamines and proinflammatory cytokines. Dysregulation is evidenced in both basal levels of system parameters - including circulating baseline levels of these hormones - as well as in patterns of dynamic response to stimuli. Chronic dysregulation is believed to confer cumulative physiological risk for disease and disability by causing damage to tissues and major organ systems.

back to top

Mechanisms and measurement

At the level of mechanisms, it is well known that a host of stressors – social and environmental elicitors of negative emotions, pathogens, physical challenges - lead to activations of physiological systems designed to maintain balance (McEwen & Stellar, 1993). Moreover, accumulating evidence supports the notion that stressors such as lower socioeconomic status, early exposure to abuse, diminished social support, and conflictual relationships (to name a few) are associated with increased risks for poor mental and physical health outcomes and mortality (Taylor, Repetti, & Seeman, 1997).

As an initial attempt to operationalize the concept of allostatic load, Seeman, McEwen, and colleagues used existing data from the MacArthur Study of Successful Aging to develop an initial measure of allostatic load that represented a simple count of the number of critical biomarkers of cardiovascular, immune, and HPA axis dysfunction on which an individual is in the highest risk quartile. The biomarkers included in this summary measure were derived from available data from the MacArthur Study that represented parameters of major regulatory systems with known or hypothesized links to various major health endpoints, including disease, disability and mortality (Seeman et al, 1997). Population-based survey research, where much attention has focused on identifying biological markers that index risk for late life disease outcomes, has provided evidence that higher levels of allostatic load are associated with increased morbidity and mortality (Seeman et al, 1997; 2001) and that higher educational attainment and better social relationships are associated with lower levels of allostatic load (Seeman et al, 2002; 2004). Alternative measurement models for allostatic load (canonical correlation, recursive partitioning) have yielded findings consistent with earlier approaches (Karlamangla et al, 2002; Gruenewald et al, 2006).

back to top

State of the evidence

Current evidence, however, does not provide definitive support for the proposed relationships between psychosocial stressors and dysregulation of biological systems supporting allostasis. Many biological regulatory systems appear to show age-related increase in levels of dysregulation, but age related changes are not uniform within populations. It is unknown what determines who exhibits such age-related changes and who manages to avoid them though a growing body of evidence points to socio-economic differences in rates of accumulation (those of lower SES exhibiting earlier and larger accumulations of AL; Crimmins et al, 2003; Seeman et al, 2004; Geronimus et al, 2006; Seeman et al, in press) as well as differences relating to socially supportive qualities of one’s relationships with others (Singer & Ryff, 1999; Seeman et al, 2002). In addition, there is a growing body of correlational evidence indicating that individuals vary in their response to stressors based on differences in personality, coping and emotional regulatory styles, and social and cultural environments (e.g., Ryff, Singer, & Love, 2004). Thus some individuals seem resilient to diseases of aging and present with a profile of positive health and well-being that may protect stress regulatory systems from dysregulation. Questions remain as to whether some people are genetically disposed to greater resilience to stress, or whether life-style, psychosocial, and socioeconomic factors are responsible for these differences.

Not all scientists are convinced that biological measures have added significantly to our understanding of risk and resilience pathways. Some have questioned whether biomarkers actually measure, as hypothesized, the mediators of psychosocial impacts on health. Even among social and behavioral researchers committed to employing biomarkers and physiological measures in their research, there is an awareness of serious gaps in the theoretical model and needs for refinement of methodologies and analytical strategies. Much more needs to be known about which biomarkers (and/or combinations of biomarkers) are most useful in predicting health outcomes in older age, and which psychosocial factors are the key predictors of change in these biomarkers over time. These researchers continue to wrest with measurement issues around “allostatic load” as they strive to keep apace of findings emerging from the biological literature on disease markers and key indicators of age-related physiological decline. The current workshop will focus on these pressing conceptual and measurement issues.

While the allostatic load model places emphasis on the biological pathways through which psychosocial factors have their effects on health outcomes, it should be emphasized that the fully integrative biopsychosocial model is recursive, with numerous and complex hypothesized bidirectional causal pathways. Real progress in integrative physiology requires the development of new analytical techniques that enable exploration of these multiple causative links, as well as advances in measurement at the level of the physical environment, the social context, the psychological subject, and the biological markers and mechanisms underlying health and disease.

back to top

Topics to be addressed in the workshop and papers

Conceptual and theoretical issues

Allostatic load is both a theoretical construct and a measure of cumulative wear and tear on physiological symptoms due to chronic stress. As a theoretical construct, it is a preliminary attempt to formulate the relationship between environmental stressors and disease, by hypothesizing mechanisms whereby multiple kinds of stressors confer risk simultaneously in multiple physiological systems (Singer & Ryff, 2001). Important questions surrounding the concept of allostatic load concern (1) the relationship of diverse research findings to the overall theory, (2) the distinction between primary mediators of disease and secondary outcomes, (3) the links between allostatic load and specific disease outcomes, i.e., does allostatic load represents a unique “syndrome” or an early stage on the pathway to multiple diseases, and (4) whether there are different allostatic load pathways, and if so, how do these differ from unique disease pathways?

back to top

Methodological: measurement of allostatic load

Which biomarkers? The currently available assortment of biomarkers for measuring components of allostatic load is extensive and includes: anthropometric, cardiovascular and metabolic measures, inflammation markers, measures of HPA axis and SNS activity, measures of renal function, lung function, bone density, immune functioning, antioxidants, genetic markers, and adiposity. One of the aims of this workshop is to encourage researchers to move beyond examination of how a single psychosocial factor influences a single biological system or health outcome and toward a more comprehensive view of the various profiles of dysregulation that may develop. Greater attention needs to be paid to primary mediators that can affect multiple other regulatory systems (e.g. cortisol, inflammatory cytokines, catecholamines) as well as to anabolic hormones that may mediate resilience (e.g. Igf-1, DHEA, testosterone).

Multi-system approaches. It is inherent in the hypothesis that a lifetime of exposure to stressors of diverse origin will have cumulative effects on multiple biological systems that we must be interested in change in multiple systems over time. A limitation of most prior work on allostatic load model is that it has been based on measures of biomarkers from only one point in time. However, if chronic stress disrupts allostatic load component systems sequentially, it may be more appropriate to look first at early alterations in neuroendocrine systems, later at elevations in inflammatory markers, and even later at markers of metabolic syndrome. One key question to be addressed by the workshop will be what the proper models are for looking at this. Longitudinal measurements of biomarkers may provide critical information about the pathways of cumulative dysregulation that lead to disease.

Biological aging as the background. The impact of chronic and daily stressors on regulatory systems occurs against a background of normal age-related declines in immune function, and the interaction of these factors may pose additional health risks. Older adults appear to show even greater immunological impairments associated with stress or depression than young adults, making factors that are associated with greater stress potentially more lethal to the elderly (Kiecolt-Glaser & Glaser, 2002). It will be important to understand how accumulation of allostatic load interacts with “normal” age-related dysregulation in each of the systems that modulate stress responses.

Multiple levels of analysis. Additionally, this workshop should explore research strategies that permit the examination of these relationships at multiple levels (from more molecular up through more macro-level assessments of social conditions) by offering models for exploring pathways between levels and examples of successful research that has adopted this approach. Importantly, measurement of biological dysregulation requires both an examination of differences in basal activity as well as and examination of system dynamics in response to challenge.

Measures of reactivity in systems. While measures of stress reactivity tend to be of higher cost, they are ultimately important for testing the concept of allostatic load, since patterns of reaction in these systems may be the earliest signs of dysregulation. Research is needed to clarify where in system dynamics dysregulations are most evident or most consequential, e.g. in response versus recovery parameters. A variety of challenge paradigms exist for studying the effects of stress on regulatory systems, including wound healing, vaccination response, and social and cognitive stress paradigms designed for use with measures of psychophysiological reactivity. What do the findings from these sorts of studies tell us about individual differences in stress responsivity, links between dysregulation and specific psychosocial factors and health outcomes, and relationships between dysregulation on the various measures? Emotional reactivity plays a pivotal role in the response to stress. How much of emotional reactivity is biologically, genetically or environmentally determined? What are the current measures of emotional reactivity to stress, and what is known about how these develop and vary over the life course?

Lab-survey linkages. Importantly, links need to be forged between studies that focus on detailed evaluations of biological mechanisms in the laboratory or field and population-based studies with greater generalizability that don’t allow for such detailed physiological assessment. Given the complementing strengths of laboratory and ambulatory research it is now possible and fruitful to extend traditional laboratory studies with experience sampling methodologies and corresponding ambulatory physiological, behavioral, and endocrinological assessments. Such studies would provide insight into the possibility that dysregulation is initially most evident in patterns of response, while more basal activity is less easily affected and may only show age-related changes relatively late in the process of development of disease and dysregulation. Participants are encouraged to envision ideal research designs that would provide insights on changes in regulatory systems and their association with past and current psychosocial factors. The workshop should provide a picture of state of the art research in this area, and point to potentials for and obstacles to improvements in research designs.

Genetics. As we learn more about the composition of the genome and mechanisms controlling gene expression, and as inexpensive methods for measuring DNA develop rapidly, it will be possible to integrate genetic data with more comprehensive models that include information on behavioral, social and other characteristics of individuals, and allowing study of gene-environment interaction. Additionally, there is an emerging interest in studying the genetic basis for individual differences in response to or susceptibility to stress, a more nuanced appreciation of the specific psychosocial phenotypes that play a role in health outcomes and the underlying genetic and gene by environment interactions that determine their expression.

back to top

Additional measurement issues include:

What selection of currently available biomarkers are most predictive of specific health outcomes, and which are indicative of health status not revealed by self-report?

What are the criteria by which new biomarkers should be evaluated for inclusion in research on cumulative physiological risk?

What cumulative measures of allostatic load show most promise? How are they best analyzed? How should measures be combined to best clarify processes of change in health?

Do different biomarkers mean different things at different times in life, and if so, which measures are meaningful at which ages?

For different regulatory systems that change over time, how does one model the interrelated effects of these systems on outcomes?

What needs to be done to clarify the links between primary mediators and secondary outcomes (e.g., inflammation, antioxidants and hormones as primary, bone health and hypertension as secondary)?

How do measures of allostatic load relate to indices such as Metabolic Syndrome and the Framingham Risk Score?

How reliable are existing biomarker measures of cumulative risk? How reliable are measures of the individual components of cumulative risk scores?

Relatedly, explorations of variability in primary mediators over time are needed, as little is known about this. Is variability on specific measures itself an outcome of potential significance? Are individuals who show greater variability at greater risk than those who are relatively stable on measures?

Are different physiological systems more or less vulnerable to different types of psychosocial stressors (financial, social isolation, interpersonal conflict) and what are the implications for disease outcomes?

Is there more than one allostatic load profile, e.g. hypocortisolism vs. hypercortisolism; hypercholesterolism vs. hypocholesterolism, and do the health consequences of these profiles change with age?

What is the relative importance of early life exposures to stress, versus later life exposures? Are there paths of no return, critical periods? Are effects of stress reversible? Does this vary by life period exposure?

How do profiles of allostatic load differ by ethnicity and gender?

How should we account for individual differences in responses to stress?

What kinds of studies could be envisioned that would permit both pre- and post-stress biological measures, as well as measures of individual difference characteristics that may modify patterns of physiological response?

back to top


Cannon, W.B. (1932). The Wisdom of the Body. New York: Norton. Cited in Schulkin J. Allostasis, Homeostasis, and the Costs of Physiological Adaptation. Oxford: Cambridge University Press, 2004

Crimmins EM, Johnson M, Hayward M, Seeman T. Age differences in allostatic load: an index of physiological dysregulation. Experimental Gerontology. July 38: 731-734, 2003.

Evans GW, Kim P, Ting AH, Tesher HB, Shannis D. Cumulative risk, maternal responsiveness and allostatic load among young adolescents. Dev Psychol 43:341-51, 2007.

Geronimus A.T., Hicken, M., Keene, D., Bound, J. "Weathering" and age patterns of allostatic load scores among Blacks and Whites in the United States. AJPH 96, 826-33, 2006

Glover DA, Stuber M, Poland RE. Allostatic Load in women with and without PTSD symptoms. Psychiatry, 69:191-203, 2006.

Gruenewald TL, Seeman TE, Ryff CD, Karlamangla AS, Singer BH. Combinations of Biomarkers Predictive of Later Life Mortality. Proceedings of the National Academy of Sciences, 103(38):14158-14163, 2006

Hu P, Wagle N, Goldman N, Weinstein M, Seeman TE. 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, Oct 20;:1-12, 2006. [Epub ahead of print].

Karlamangla, A.S., B.H. Singer, B.S. McEwen, J.W. Rowe, and Seeman TE. Allostatic lead as a predictor of functional decline: MacArthur Studies of Successful Aging. Journal of Clinical Epidemiology, Vol 55, No. 7, 696-710, July 2002.

Kiecolt-Glaser JK, Glaser R. Depression and immune function: central pathways to morbidity and mortality. Journal of Psychosomatic Research, 2002 Oct;53(4):873-6.

Kubzansky LD, Kawachi I, Sparrow D. Socioeconomic status, hostility, and risk factor clustering in the Normative Aging Study: any help from the concept of allostatic load? Annals of Behavioral Medicine 1999;21(4):330-8

McEwen BS. Protective and damaging effects of stress mediators. New England Journal of Medicine 1998;338:171-9

McEwen BS, Seeman TE. Protective and damaging effects of mediators of stress. In Socioeconomic Status and Health in Industrial Nations: Social, Psychological and Biological Pathways, Adler NE, Marmot M, McEwen BS (eds), NY: NY Academic of Sciences, 896:30-47, 1999.

McEwen BS, Stellar E. Stress and the individual: mechanisms leading to disease. Archives of Internal Medicine 1993;153:2093-101.

Ryff CD & Singer B (eds), Emotion, Social Relationships and Health. New York: Oxford Press, pp189-209, 2001.

Ryff CD, Singer BH, Dienberg Love G. Positive health: Connecting well-being with biology. Philos Trans R Soc Lond B Biol Sci. 359:1383-94, 2004.

Schulkin J. Allostasis, Homeostasis, and the Costs of Physiological Adaptation. Oxford: Cambridge University Press, 2004

Seeman TE, Singer B, Horwitz R, McEwen BS. "The Price of Adaptation --Allostatic Load & Its Health Consequences: MacArthur Studies of Successful Aging" Archives of Internal Medicine 157:2259-2268, 1997.

Seeman TE, Singer B, Rowe J, McEwen B. Exploring a new concept of cumulative biological risk -- Allostatic load & its health consequences: MacArthur Studies of Successful Aging. Proc Nat Acad Sci USA 98(8): 4770-4775, 2001.

Seeman TE, Singer B, Ryff C Levy-Storms L. Psychosocial factors and the development of allostatic load. Psychosomatic Medicine 64:395-406, May/June 2002.

Seeman TE, Crimmins E, Bucur A., Huang MH, Singer B, Bucur A, Gruenewald T, Berkman LF, Reuben DB. Cumulative Biological Risk and Socio-Economic Differences in Mortality: MacArthur Studies of Successful Aging. Soc Sci & Med 58, 1985-1997, 2004

Seeman TE, Merkin S, Crimmins E, Koretz B, Charrette S, Karlamangla A. Education, Income and Ethnic Differences in Cumulative Biological Risk Profiles in a National Sample of US Adults: NHANES III; Social Science and Medicine, in press.

Selye, H. (1956). The Stress of Life. New York: McGraw-Hill. Cited in Schulkin J. Allostasis, Homeostasis, and the Costs of Physiological Adaptation. Oxford: Cambridge University Press, 2004

Selye, H. (1974). Stress without Distres. New York: New American Library. Cited in Schulkin J. Allostasis, Homeostasis, and the Costs of Physiological Adaptation. Oxford: Cambridge University Press, 2004

Shannon M, King TL, Kennedy HP. Allostasis: a theoretical framework for understanding and evaluating prenatal health outcomes. J. Obstet Gynecol Neonatal Nurs. 36:125-134, 2007.

Singer B, Ryff C, Seeman TE. Operationalizing Allostatic Load. In Allostasis, Homeostasis, and the Cost of Physiological Adaptation. J. Schulkin (ed). Pp 113-149, 2004

Singer B, Ryff CD. Hierarchies of life histories and associated health risks. Annals of New York Academy of Sciences 1999;896:96-115.

Taylor SE, Repetti RL, Seeman TE. "What is an Unhealthy Environment and How Does It Get Under the Skin?" Annual Review of Psychology, 48:411-47, 1997

back to top

Charge to Participants in the NIA Workshop on Allostatic Load

In preparation for the upcoming NIA Exploratory Workshop on Allostatic Load, participants were asked to prepare a short (2-4 pages) statements outlining their views on how research on allostatic load and the study of cumulative risk more generally can be most fruitfully advanced in the behavioral and social sciences. These statements are intended to set the foundations for dialogue, at the workshop, on research and resource needs for achieving this goal.

As described in the foregoing background document, there are a number of conceptual and methodological issues that could be addressed at the workshop. Participants were asked to offer creative, well-informed input on where the emerging research opportunities and needs lie within thier own and related fields.

Participants were asked to consider the following when preparing their statements:

  • How has the concept of allostatic load had a positive (or negative) impact on work in your field? Are there more valuable alternative approaches to the study of cumulative physiological risk, and if so, what are they and what is their relative advantage?
  • What could be done to make this concept more valuable to research on social and behavioral factors in aging?
  • What are the most critical measurement issues, in your view? Are there specific data needs for addressing these questions?
  • What would be the most important "next steps" in moving research on allostatic load, or cumulative physiological risk more generally, forward?

Statements from invited participants follow.

John Cacioppo

University of Chicago

What is adaptive or maladaptive, healthy or unhealthy, depends on context, and what may be good for one tissue may be life saving, but may have negative impact on another tissue with mortal consequences in the long run. The regional positioning of immunocytes in response to acute stress may provide the host with a selective advantage should aggressive behavioral interactions lead to cutaneous wounding and the possibility of infection (Dhabhar & McEwen, 1997). The selective advantage that may accompany acute stress does not extend to chronic forms of stress, however, as the prolonged activation of the HPA axis and sympathetic nervous system seen in chronic stress tends to suppress cellular immunity (Lupien & McEwen, 1997; Sheridan, 1998), reduce response to vaccination (Kiecolt-Glaser et al., 1996), and slow the healing of experimental cutaneous and mucosal wounds (Kiecolt-Glaser et al., 1995; Marucha et al., 1998; Padgett, Marucha, & Sheridan, 1998). These complexities underscore the interdependence between organisms and their physical and social environments. Restorative mechanisms such as a balanced diet, moderate exercise, sleep, and rich social connections may have the salutogenic (e.g., stress buffering) effects, whereas a diet high in fat, the use of tobacco and alcohol, a sedentary lifestyle, and hostility and isolation exacerbate the deleterious effects of chronic stress. Our understanding of the complex regulative and restorative processes of the organism – and the balance within –is therefore fostered by a multi-level integrative analysis.
Although the biological substrates for the etiology and course of chronic disease are influenced profoundly by the physical and social world, it remains to be determined to what extent these influences result in individual differences in allostatic load as a consequence of differential exposure to stressors, differential reactivity to stressors, or differential recovery from stressors – and why.
Second, a good deal is now known about stress physiology (e.g., Chrousos, 2001), but the rostral neurobehavioral systems that orchestrate organismic-environmental transactions (Berntson et al., 1998, in press) and the psychological transduction mechanisms (e.g., health behaviors, health utilization, stress buffering) remain only partially mapped. "Stress" has been assigned a special role in the development of allostatic load, but the concept of stress itself is often vaguely (or circularly) defined. Operationalizations and measures across studies, especially across animal and human studies, were regularly so different (e.g., restraint, hypoglycemic, orthostatic, mathematic stressors) that results are sometimes difficult to compare or reconcile.
Third, stressors are not always negative, either, as positive as well as negative events are considered stressors in studies focused on predicting health (Holmes & Rahe, 1967). Further complicating matters, the measurements of stress within a given study are often so weakly correlated that they provide poor convergent validity for the construct of stress (e.g., Lacey, 1959; Johnson & Anderson, 1990). In short, neither stress nor health is a simple, unitary concept, and the search for a singular universal mechanism relating stress to health is doomed to failure. The concept of allostatic load, too, while useful at a molar level of analysis, is misleading if applied to specific underlying mechanisms as if there were a single cause of wear and tear. The concept is useful because it represents a broad and multifarious category of specific and largely unrelated transduction mechanisms that contribute to the wear and tear on the organism.

Why not simply ignore the molar constructs of integrative physiology and focus on the details of the cellular machinery? As Claude Bernard (1878/1974) opined over a century ago and Lewontin (2000) has echoed more recently, the organization and function of the elemental parts of an organism can be understood comprehensively only within the context of its transactions with its physical and social environment. Lewontin’s (2000) analysis suggesting that the reduction in mortality from infectious diseases during the late 1800’s and early 1900’s is attributable to a general trend of increases in real wage, an increase in the state of nutrition, and a decrease in the number of hours worked – that is, a decrease in the wear and tear on the organism and improved opportunities for the organism to heal itself – hints at causal factors and targets of interventions to which we would be blind by focusing on cellular mechanisms alone. Although we are beginning to understand how the brain integrates the regulatory and restorative forces of the body to foster health and adaptation to environmental challenges, it is clear that health psychology will have much to contribute to this understanding for a long time to come.
The theory of allostasis remains to be fully delineated. Improvements in operationalizations and construct-valid measures would promote theoretical tests and refinements. Specification of the neurobiological basis of allostasis would advance our understanding of the functional system. Collaborative studies involving animal models and longitudinal human studies would promote a mapping of CNS-PNS orchestration and the decay of this orchestration.

Sheldon Cohen

Carnegie Mellon University

Allostatic Load/Cumulative physiological risk

This is a quick list of my thoughts about Allostatic Load (AL) Theory and primarily about AL measurement. I should say that while I have been involved multiple discussion about AL that occurred in the MacArthur SES network and in various other forums starting when Bruce first drafted the NEJM paper, I do not carefully follow research in AL or changes in its conceptualization or measurement. Hence I hope you will excuse any issues raised below that may have been addressed more recently as the theory and measurement of AL has grown.

General Opinion: I like the AL heuristic, I don’t like the way it has been operationalized. McEwen’s NEJM paper suggests that enduring chronic stressors can result in the disturbance (dysregulation) of multiple systems. Dysregulation can mean that the system overreacts, underreacts, overreacts but cannot return to normal, etc. The application of this idea to the study of specific systems can be powerful. For example, chronic stress influences how effectively we regulate our immune response, it does not just cause immunosuppression.

Which system is influenced in an individual presumable would be moderated by individual vulnerabilities, e.g., genetic, previous illness, etc. Less clear (and in my recall not a part of the original theory) is the hypothesis that chronic stress disrupts multiple systems and consequently aggregating across systems is an appropriate way of measuring AL. However, this is how AL is generally measured.

I think there have been two underlying weakness in AL measurement. The first, as suggested above, is aggregation across multiple systems without adequate consideration of the possibility that a single or specific set of individual systems are what is important. The second is that it is based on what has been (and in some cases can easily be) measured rather than what needs to be measured to provide an assessment of dysregulation. (This includes CARDIA at 15 years where pretty much every standard measure that might assess AL and was practical was selected, rather than asking what is the minimum requirement for assessing dysregulation across multiple systems.)

Up or Down or in the Middle? Clearly a problem with many attempts to assess AL is that they don’t focus on regulation but rather on markers moving in what is considered a direction of pathogenic risk based on standard risk factor type analysis (e.g., high blood pressure is bad). A measure consistent with the theory would focus on dysregulation whether assessed as low, high, over and under reaction to stress, or inability to return baseline after stress.

Apples and Oranges? Many of the studies presumed to measure "allostatic load" aggregate across biomarkers representing a range of different physiological systems. Interestingly, they generally unit weight biomarkers, not systems. Many of the biomarkers are measures of various cardiovascular functions and metabolic functions. As a consequence, if the measure is related to something (e.g., measure of SES or chronic stress or disease risk) it could be entirely due to its influence on a single system. If an association is not found, there may be relationships with individual systems, but the overall effect is wiped out (or the effect size is reduced) by randomness across other systems.

If one believes that aggregating across systems is important, than one should 1) define the important systems to measure (e.g., ANS, SAM, HPA, Cardiovascular, Immune, etc.); 2) develop adequate measures of the REGULATION of each of these systems; and 3) equally weigh systems or determine weights based on a theoretical basis in creating an aggregate.

Specificity or generality? There are reasons to think that specific biomarkers used in assessing AL should be associated with specific disease outcomes. (For example, cardiovascular system markers are associated with cardiovascular disease risk). Comparisons of additional variance accounted for after controlling for the system that one expects to predict would be helpful. For example, it is possible that failure in one system drives (or is otherwise correlated with) failure in another. This doesn’t mean that you need failure in both to influence a specific disease outcome, although aggregation of system failure may make it look this way.

Does it depend when it happens for whether it counts toward AL? AL measures also tend to include both what McEwen called primary and secondary mediators. Primary measures are immediate nonspecific biomarkers of stress, e.g., catecholamines and cortisol. Anything further downstream is considered secondary. Current work does not differentiate between secondary measures that are markers of disease risk versus disease pathology.

More recently, inflammation has been considered part of the primary mechanism (see background statement). Although inflammation may play an upstream role in disease progression (as hypothesized for coronary artery disease), it clearly is a measure of underlying inflammatory processes. Inflammation occurs in response to tissue damage. Therefore it is a marker of an ongoing (although sometimes premorbid) disease process. Hard to see it as a primary mediator.

Should dysregulation be assessed as a process?
Maybe we should consider an assessment of AL that addresses HOW FAR DOWN STREAM the dysregulation has moved. For example, the difference between a person with elevated cortisol and one with elevated cortisol and inflammatory markers?


  1. Develop the measures as assessments of dysregulation (not just convenient measures of risk). This would include some thought about the nature of dysregulation and assessments of reactivity and recovery as well as the usual basal measures.
  1. Develop measures of individual systems. This requires a clear conceptualization of each system and preliminary empirical work to see how measures cluster. For example, Steve Cole did some initial work with measuring the ANS that was successful in linking social inhibition to HIV progression.
  1. Consider the individual variables that might predict individual differences in what system(s) would show vulnerabilities under stress.
  1. Consider an assessment of AL that addresses HOW FAR DOWN STREAM the dysregulation has moved. For example, the difference between a person with elevated cortisol and one with elevated cortisol and high blood pressure, and one with elevated cortisol, high blood pressure and inflammatory markers.
  1. Studies that focus on the stability/changes in the stress (social and physical) environment over time.

Steve Cole



I study social regulation of gene expression. My work maps the ksocial signal transduction pathways" by which socio-environmental processes regulate broad patterns of gene transcription across the entire human genome (and across viral and tumor genomes). This work is generally molecular biological and computational/bioinformatic, with a concrete focus on neuroendocrine receptors, intracellular signal transduction pathways, transcription factor activation, and other aspects of gene regulation. My lab focuses on the subset of gene regulation pathways that are empirically responsive to social inequality, social isolation, and "stress". We are not concerned with the activity of any individual gene, except insofar as it provides a good generalizable example of how social regulation of gene expression takes place. Our primary objective is understanding the higher logic of a socio-environmentally responsive genome, including:

1) Which broad groups of genes are sensitive to social factors?,
2) Which signal transduction pathways mediate those effects?,
3) Which genetic polymorphisms moderate those effects?, and,
4) What teleologic purpose is served by socially responsive genome?

I was trained as a social psychologist (Stanford), am employed as a molecular biologist (UCLA School of Medicine and HopeLab Foundation), and do most of my work with math and computers. I began this work as a virologist, detailing the molecular processes by which psychosocial factors drive the HIV-1 virus. We eventually determined that HIV-1, and most other successful human viruses, have evolved to take advantage of a biochemical "stress niche" in their hosts (us). My work now focuses on understanding that human genomic niche, and ameliorating its impact on viral infections and cancer.

(1) How has the concept of allostatic load had a positive (or negative) impact on work in your field? Are there more valuable alternative approaches to the study of cumulative physiological risk, and if so, what are they and what is their relative advantage?

I believe the general theory of Allostatic Load is likely correct, potentially quite useful, and not currently being operationalized in ways that are scientifically helpful. My inclination would not be to seek other theories, but to seek better biological (and behavioral) measures of cumulative biological "wear and tear" and health risk. It would be helpful if researchers maintained a clearer distinction between the hypothesized mediator (allostatic load) and the outcomes it is invoked to explain (e.g., disease biology).

(2) What could be done to make this concept more valuable to research on social and behavioral factors in aging?

The concept of allostatic load is valuable, but most biomarkers currently used as measures of allostatic load do not fit the bill. Many biomarkers offered as measures of allostatic load may actually be measures of the chronic disease processes that allostatic load is invoked to explain (e.g., cytokine levels) or measures of transient states rather than cumulative states (e.g., hormone levels). Telomere length is, in principle, a good biomarker of cumulative challenge because the state of the biological "analyte" (the thing being measured) is known to change in a cumulative fashion with stimulation (i.e., cell turn-over progressively shortens the terminal chromosome cap, though telomerase introduces complications). Glucocorticoid resistance in leukocytes may also reflect cumulative exposure, and it is conceivable that flattened diurnal cortisol slopes might also (more longitudinal work is needed to validate the later). Current hormone or cytokine levels do not have that cumulative property and thus cannot serve as measures of a cumulative process. Any biological parameter that does not show a monotonic change in value over time must indicate something else instead of (or in addition to) cumulative biological challenge.

(3) What are the most critical measurement issues, in your view? Are there specific data needs for addressing these questions?

a) Better biomarkers of cumulative biological effects,
b) better behavioral/neurobiological measures of cumulative risk exposure, and
c) a clear distinction between measures of cumulative challenge and measures of current disease biology.

Neural proxies for cumulative challenge should be developed more fully (e.g., functional neural imaging responses to targeted probe stimuli, progressive innervation or denervation of solid tissues, epigenetic regulation of neural genomes or neurally-responsive cells such as leukocytes, etc.). Much of this biological "history" may be encoded in the internal regulatory structure of cellular response to environmental stimulus (e.g., glucocorticoid resistance) rather than in the extracellular parameters released from cells (e.g., circulating hormone levels).

(4) What would be the most important "next steps" in moving research on allostatic load, or cumulative physiological risk more generally, forward?

a) A greater focus on biological parameters that are truly cumulative (e.g., more the quality of telomere shortening) or are functionally historical (e.g., glucocorticoid resistance)
b) treatment of "state" parameters (e.g., hormone levels, cytokine levels, CRP levels, etc.) as outcomes to be explained rather than cumulative mediators, and
c) a greater understanding of cumulative neurobiological allostasis, and its biomarkers. This would provide measurement advantages (relative to self-report) and fill in an important missing piece in the basic theory of how historical environmental stimuli alter the current functional activity of the body.

I believe the later concept represents the most significant contribution of Allostatic Load theory. How is it that past events can affect the functional characteristics of the present body? To answer this question, it might be helpful if Allostatic Load theorists borrow more deeply and directly from developmental biology and psychology. Telomeres are clever in this regard. What else might we borrow from the biology of life-span development?

Eileen Crimmins

University of Southern California

Re: Relevance of Allostatic Load to Measuring Risk in Large Populations
How has the concept of allostatic load had a positive (or negative) impact on work in your field? Are there more valuable alternative approaches to the study of cumulative physiological risk, and if so, what are they and what is their relative advantage?
The work on allostatic load has been central in promoting the idea that the body has a set if integrated physiological systems. The demonstration that large components of allostatic load can be measured in community settings, has led to the incorporation of many of these indicators in population level surveys. All of this has been extremely valuable and changed the world of data collection very rapidly
However, the measurement to date still has limited physiological coverage in populations because of difficulty of measuring some physiological systems. So the theoretical idea may not be fully addressed. In addition while allostatic load has an emphasis on reaction to stress, there is a category of markers of organ reserve that might be more appropriate for use as an indicator of "frailty" or loss of function that is appropriate for evaluating population health change. In general a separation into indictors of early and late trajectory physiological change would be useful.
What could be done to make this concept more valuable to research on social and behavioral factors in aging?
Review of conceptual issues to clarify how empirical tests are able to address theoretical ideas.of allostatic load and integration of this concept with ideas of population health change.
What are the most critical measurement issues, in your view? Are there specific data needs for addressing these questions?
In population studies, the use of new methods has increased the data collection more rapidly than the development of assays using non invasive collection methods such as saliva and blood spots.
What would be the most important "next steps" in moving research on allostatic load, or cumulative physiological risk more generally, forward?
Most immediately, the large national samples are not measuring stress indicators. This seems essential to test the most important hypotheses about stress and discrimination as factors in health differentials.

Elissa Epel


Influence of Allostatic load (AL), Allostasis and health.
The concept of allostasis and allostatic load have had a large impact on my research. AL model promotes a developmental model of disease progression, toward general pathways, across systems, away from single biomarkers. It is important in stress mechanism research both to understand general pathways of chronic disease (AL, met syndrome) as well as specific pathways (ie, Lutgendorf’s model of ovarian cancer, Dhabhar’s model of skin cancer).

To understand health, not disease, it is helpful to measure allostasis (albeit a moving target) and, further, whether ‘enhanced allostasis’ exists (functioning above baseline, characteristic of youthful systems) (See Bower et al, 2007). Examples of enhanced allostasis may be the high levels of variability (heart rate variability, hormonal pulses) in youthful systems, which degrade over time. The concept of recovery is crucial to understanding allostasis. Most reactivity studies do not examine longer term recovery from acute stressors. This may be one critical period during which emotional regulation and cognition (positive appraisals vs. rumination) are more closely tied to physiology/states of arousal. My preliminary data (that I will show) suggests that emotional reactivity and rumination during recovery are linked to cellular aging markers. High variability and rapid recovery may represent enhanced allostasis, which may be predictors of longevity. This is mere speculation, but to understand aging, we need studies that test whether there are unique mechanisms that predict longevity, which may be different than those that predict early disease. The AL model provides hypotheses or phenotypes about healthy vs. aged profiles of arousal during acute and chronic stress states. The concept is clarifying, the operationalization of both allostasis and AL is challenging (below).

What could be done to make this concept more valuable to research on social and behavioral factors in aging?

It would be helpful to have better measurement of primary mediators, which is challenging. We need better identification of the key measures in an allostatic load battery, and which aspects are most closely tied to behavior, stress, and genetics. One long standing obstacle in stress research, only recognized more recently, is that primary mediators such as cortisol levels are a dynamic measure, and dysregulation can lead to both hyper or hypocortisolemia. Researchers need better statistical models to deal with ‘bimodal’ variables of dysregulation, and to use variance like this as a way to understand different phenotypes of dysregulation. We need markers of chronic stress that don’t depend on how stressful one’s day was. For example, imaging measures of adrenal volume, thymus involution, and possibly dexamethasone nonsupression, should represent a history of chronic stress arousal, at least ACTH or cortisol.

It would be helpful to develop better measurement of chronic stress. One might wonder why AL is not related to distress in several population based studies. Often we are trying to link a ‘state’ measure available, like depression over the last two weeks with what may be largely a long term reflection of lifestyle and wear and tear. Why should recent mood be strongly linked to long standing footprints of dysregulation? We need cumulative models of distress. We are currently trying to measure lifetime histories of depression better. Measuring stress retrospectively is even more problematic. Measurement of emotional reactivity may be more closely linked with allostasis than trait measures of negative affect or depression. What types of stress are most damaging, and what does this look like in the brain? Studies linking brain chemistry and activity with peripheral biomarkers would be invaluable. However, we are limited by our inability to measure important central peptides in humans peripherally (eg, NPY, oxtytocin, opioids in the VTA) and there is a lack of ligands that would allow us to see PET scan activity.

Given the central role that chronic stress plays in accelerating aging, we need a better multilevel understanding of stress vulnerability and stress resistance. The problem becomes obvious when we realize that there is no agreed upon operationalization of either term. What is stress vulnerability or stress resistance, psychologically or physiologically? Many of us have answers to that, from our individual paradigms, but they probably have little "shared variance." It is imperative to identify the most common phenotypes of a maladaptive stress response—including whatever components are most important in the pathway toward aging--- affective, cognitive, genetic, and in terms of neural activity and biochemical substrates centrally and peripherally.

We need to move beyond tautological models – eg, linking high blood pressure to hypertension to CVD, to studies that point to mechanism. This is where models of somatic cellular aging or immunosenescence, and gene expression may be helpful. Focusing on cell based markers of biological aging allows us to measure aging in youthful systems as they develop, and allows examination of life course perspectives. Markers of how cells age –such as telomere length—and the biochemical mediators such as Insulin exposure, oxidative stress, and inflammation-- may reflect poor allostasis even in young people. However, mid life is a time that a critical mass of mitotic cells become senescent and thus biological aging becomes manifested in phenotypes of aging (graying, wrinkling, clinically measurable disease). It is helpful to have aging studies start at mid-life or earlier.

What would be the most important "next steps" in moving research on allostatic load, or cumulative physiological risk more generally, forward?

There is a need for ‘mechanistic’ longitudinal studies. Given the nonlinearity of physiological processes (eg, bidirectional relationships, webs of interconnections), cross sectional research can lead to false assumptions. With limited resources, there should be more focus on causal mechanisms. This includes longitudinal studies with experimental studies embedded (like MIDUS). Well characterized samples could be taken advantage of. The lack of investment in samples that cannot be followed sufficiently, due to limited funds to extend research longitudinally, is an obstacle to understanding the process of aging.

Once we have developed more advanced conceptual models of phenotypes of "stress" –we need multilevel research examining how "stress vulnerability" leads to transmission of disease risk and conversely, how "stress resilience" may confer longevity. This might require multigenerational studies that include prenatal conditions. Large longitudinal studies that examine women and men in young adulthood, and how they pass on traits such as emotional reactivity, stress hormone reactivity, insulin resistance, and telomere length to their progeny, and substudies to examine mechanisms in depth. Such studies require multi level analyses, and need to include social factors, genetic, behavioral phenotypes, neurobiology at least as well as it can be measured in humans.

Importance of health behaviors as moderators and mediators not covariates.
How much of AL is related to health behaviors, and how much of our health behaviors are environmentally constrained or influenced by genetic variations? Linear models are limiting, and the large effects of stress may be in interaction with behavior. Health behaviors are often used as covariates but actually affected by stress. Behavior can mediate negative effects of stress (such as the interaction of high fat diet and stress) or, in the case of exercise, can moderate and buffer impact of stress. As an example, there have been many iterations of human studies addressing relationships between stress and visceral fat in cross sectional ways. Some studies find this relationship and others do not. New rat research shows how chronic stress interacts with a high fat diet to recruit and enlarge visceral fat cells through NPY. Chronic stress alone did not significantly increase visceral fat (ie, no main effects), but only in interaction with eating fat. The action was in the interaction, and it required an animal studies to unravel the mechanism. We need more animal studies addressing questions about behavior/stress interactions.

Lisa Fredman

Boston University

  1. How has the concept of allostatic load had a positive (or negative) impact on work in your field? Are there more valuable alternative approaches to the study of cumulative physiological risk and if so, what are they and what is their relative advantage?

I am a psychosocial epidemiologist and my research focuses on the health effects of caregiving stress. I think that the concept of allostatic load has had a big impact, both positive and negative, on the field of psychosocial epidemiology, and little impact on caregiving research. It has had a huge impact on my own research: here is how.

I have been conducting a prospective cohort study on a sample of elderly women caregivers and non-caregivers who are participants in the Study of Osteoporotic Fractures (aka Caregiver-SOF). Its theoretical framework is the stress and coping model. The Caregiver-SOF caregivers are more stressed than the non-caregivers, but have lower rates of functional decline and mortality. I could not explain this methodologically or theoretically until I read Teresa’s studies of allostatic load. The strengths of the concept were that the "wear and tear of chronic stress" affected multiple, inter-related homeostatic systems, and that it predicted established measures of health decline in an epidemiologic study. High allostatic load could be the physiologic pathway between caregiving and health decline. I performed a cross-sectional pilot study: caregivers and non-caregivers did not differ on allostatic load, although the caregivers had more stress. Therefore, I theorized that a subset of caregivers – those who were stressed AND developed high allostatic load – would have higher rates of health decline, whereas those who were stressed but who did not progress to allostatic load would not differ from non-caregivers. The prospective study design would reveal the temporal relationships between chronic caregiving stress, development of allostatic load, and health decline.

I submitted and resubmitted several proposals to test this hypothesis to NIH. Reviewers from multiple study sections criticized the concept and the measurement of allostatic load on the grounds that a single score was used to reflect a multidimensional concept, validity studies were lacking, and that the measure combined cardiovascular, metabolic, and neuroendocrine indicators which could mask the true physiologic mechanism.

Based on these critiques and recent studies of stress-induced metabolic syndrome, I revised my conceptual model. Metabolic syndrome replaced allostatic load as the pivotal mediator between chronic caregiving stress and health decline. This new model incorporates theories that stress-induced disruptions in the neuroendocrine and/or immune systems may increase the risk of metabolic syndrome, leading to health decline. It is supported by epidemiologic studies (see Yaffe et al, 2004 on the combined effects of inflammatory markers and metabolic syndrome on cognitive decline). It is consistent with the theory of allostatic load. Moreover, data from Caregiver-SOF and our pilot study support this model.

I revised my proposals, submitted them to NIH, and got funded! That is how allostatic load has influenced my research.

In general, I think that the concept of allostatic load can give us tremendous insights into the physiological mechanism(s) by which chronic stress affects health. It would strengthen studies of caregiving outcomes which, with few exceptions (e.g., studies by Grant and by Vitaliano), focus on single biomarkers if they include biomarkers at all, and do not address effects on multiple physiological systems, or the downstream effects on physical or cognitive health decline.

  1. What could be done to make this concept more valuable to research on social and behavioral factors on aging?

In epidemiologic research, the concept of allostatic load might be more valuable if prospective studies were designed to evaluate sequential changes in the components of allostatic load, and the subsequent effects on functional, cognitive, and disease outcomes. This would involve taking multiple measures of the components of allostatic load at multiple time points, such as annual or biennial followup interviews. Such studies would allow researchers to disaggregate the components of allostatic load, and to tie it to other theories of physiological responses to chronic stress. A limitation of many epidemiologic studies is that biomarkers of allostatic load are collected at only one or two time points, which prevents seeing how changes in one system lead to downstream changes in other, multiple systems.

Secondly, since allostatic load is theorized to result from chronic stress, we need prospective cohort studies on populations exposed to chronic stress and comparable populations that are not exposed to that stress, to determine how chronic stress leads to allostatic load and components of allostatic load. Samples of caregivers and non-caregivers could serve this purpose, but the selection of non-caregiver comparison groups can introduce potential selection biases (ie, married non-caregivers are often healthier than caregivers, elderly non-caregivers are less physically active than caregivers). Other high-stress populations might include workers in high-stress occupations or spouses of veterans. I am concerned that some epidemiologic measures of "life course stressors" actually indicate non-stress risk factors for components of allostatic load: for example, low socioeconomic status may lead to metabolic syndrome through poor diet rather than through stress, so epidemiologic studies need to include measures of these confounders.

Third, perform research on factors that may prevent or reduce allostatic load components in the face of chronic stress. For example, studies could evaluate how stress reduction programs affect allostatic load components and ultimately prevent chronic stress from progressing to allostatic load.

  1. What are the most critical measurement issues? Are there specific data needs for addressing these questions?

Epidemiologists express the need for validity studies of allostatic load. In addition, there are questions about how to operationalize an "allostatic load" score. There are also questions about how to operationalize the components of allostatic load: as continuous variables, z-scores, or categorical measures. One concern is that studies that use categorical measures tend to base the cutpoint on the sample-based distribution (for example, top tertile of cortisol). But the cutpoints differ from study to study. Therefore, it would be helpful to compare measures of allostatic load and the distribution of its components in different samples and across multiple time points.

  1. What would be the most important "next steps" in moving forward research on allostatic load or cumulative physiological risk?

Researchers are increasingly including biomarkers in their studies, often without a strong conceptual framework. Perhaps research would be moved forward if there were more articles in discipline-specific journals that emphasized the conceptual strengths of the allostatic load model or on models of cumulative risk.

More research is needed on the points raised in the previous sections: the temporal relationships among the components of allostatic load; evaluating whether there are critical aspects of allostatic load, such as metabolic syndrome, that link chronic stress to disease outcomes; conducting parallel studies of allostatic load in different populations and high-stress groups to identify similarities and differences in the mediating role of allostatic load between chronic stress and health decline.

A practical issue is how to get studies on allostatic load favorably reviewed by NIH study sections. This is important for advancing research in this area.

Noreen Goldman


For the past eight years, I have been involved in a data collection effort in Taiwan known as SEBAS. This survey, which took place in 2000 and 2006, includes home-based interviews, collection of blood and 12-hour urine samples, and physicians’ health exams, from about 1000 middle-aged and elderly respondents who comprise a random subsample from an ongoing, national longitudinal survey. The underlying objective of SEBAS has been to examine the relationships among the social environment, stressful experience and mental and physical health and to identify the intervening physiological pathways. A subsidiary goal has been to evaluate the utility of the allostatic load framework and existing measures of allostatic load in elucidating the linkages between stressful experience and health.

In the Princeton workshop last spring, I outlined five sets of questions that have plagued analyses, including our own, that attempt to analyze these relationships in broad (non-clinical) populations.

  1. How well have we measured (or can we measure) stressful experience in interview surveys? How should we measure stressful experience – e.g., traumatic events, major life events, short-term stressors, perceptions of stress?
  2. Are we obtaining the "correct" biomarkers from large-scale surveys? How do we determine whether a biomarker is a potentially important component of the allostatic load framework? How often should we attempt to measure these biomarkers? How can we best parameterize a large number of biomarker values in terms of variables or scores?
  3. How well have we measured (or can we measure) individual characteristics that may moderate the impact of stressful experience (e.g., personality, coping mechanisms, social position)? Can our statistical models and sample sizes support this type of complexity?
  4. Are studies such as SEBAS focusing on too old ages – e.g., a period in the life span when the effects of social factors and challenge may be less important than earlier in life and when many persons at risk have already died?
  5. How much variation in these relationships do we expect to occur across populations (or across individuals within populations)? For example, is it reasonable to think that associations found in the US or other Western populations would be of a similar magnitude to those in Taiwan?

Although I continue to believe that these are pressing issues, I think it would be useful to focus some of our discussion on a more basic question:

What types of findings would offer strong support (or a strong refutation) of the construct of allostatic load (in human populations)?

I believe that some of the claims in the literature that purport to find support for this construct are unfounded. For example, much of the research to date on human populations that assesses the allostatic load framework does so by examining the connection between an array of biomarkers and a set of health (or survival) outcomes. Indeed, some recent studies have included in an "allostatic load score" physiological measurements that do not appear to be related to the stress response (e.g., grip strength), but are known to be strong predictors of future health status. Given the established associations between most of the biomarkers used in population surveys and a multitude of mental and physical illnesses, statistically significant associations between the two should come as no surprise. That is, the links between these physiological measurements and health may arise from numerous pathways, many of which are probably not what we would think of as stress-related mechanisms. Similarly, the demonstration that a score of allostatic load (e.g., a summation of the number of these biomarkers that fall outside normal operating ranges) has a stronger correlation with downstream health measures than do individual markers is also an expected result in light of potential measurement error and the fact that the sum of small effects is typically a larger effect.

Thus, it seems that one needs to set a higher bar to evaluate the construct of allostatic load, one that demonstrates linkages between stressful experiences and poor physiological profiles (however defined). There are two issues that require attention in this regard. One pertains to how we define a "stressor." Physiologists often focus on health-related mechanisms such as pain, pathogens, or lack of sleep. Social scientists are apt to think instead of life challenges, such as the death of a loved one, trauma, serious illness, loss of job, financial hardships, discrimination, or repeated daily hassles. Perhaps we need to distinguish here between the animal and human literatures. The second point pertaining to the link between stressors and poor physiological profiles concerns the role of socioeconomic status (SES). Social epidemiologists have argued that the social gradients in health come about in large part because persons of lower SES experience greater stress than their higher status counterparts. This conclusion is sometimes based on little more than statistically significant associations between SES (typically measured by education, income or occupational status) and a handful of biomarkers (e.g., blood pressure, cholesterol level, BMI, or fibrinogen). Much more compelling evidence is needed to identify the importance of "stress" in driving social inequalities in health. Researchers also need to keep in mind that "stress" can play two different roles in the SES-health link: persons of lower SES may experience more (or more serious) stressful events and they may be more vulnerable (have a greater physiological response) to stressors. Researchers have rarely been explicit about these two mechanisms and have generally not attempted to establish their veracity.

Igor Grant


1. How has the concept of allostatic load had an impact on work in your field? Are there more valuable alternative approaches to the study of cumulative physiological risk, and if so, what are they and what is their relative advantage?

The concepts of allostasis and allostatic load derive from a long tradition stretching back to Claude Bernard’s observations on the physiological necessity of the "fixite du milieu interieur" which was then elaborated by Walter Cannon in the concept of homeostasis. To me, the concepts of allostasis and allostatic load have been particularly useful in their ability to unite notions of homeostatsis and acute and chronic stress. Of particular heuristic importance has been the elaboration of the notion of primary mediators (e.g., catecholamines and glucocorticoids) and the categorization as effects of these mediators into primary (cellular), secondary (e.g., downstream metabolomic changes), and tertiary (actual organ disease).

These notions have been very useful in framing our own program of research into the health effects on elderly spousal caregivers of being married to and looking after a dementia husband or wife. Our overarching concept is that the stresses associated with caregiving, which are both chronic and punctuated by acute events, can lead to a state of sympathoadrenal medullary (SAM) arousal characterized by greater fluxes in circulating catecholamines, including norepinephrine and epinephrine. Consistent with notions of allostasis, we measure not only basal and stress-induced catecholamines, but also the receptor system upon which they act (in our case we focus on beta adrenergic receptors and on lymphocytes as a model of effects that may be occurring in other tissues). We then attempt to link these changes to changes in molecules such as D-dimer and IL6 (being examples of indicators of the coagulation environment and inflammation), which would be examples of secondary outcomes in the McEwen et al., model. Our work is now examining more downstream outcomes which may be thought of as transitions from secondary to tertiary outcomes — for example, changes in baroreflex, in endothelial function as measured by flow mediated dilation, and evidence of early arteriosclerosis as determined through ultrasound.

Our model also examines the effects of resiliency factors such as coping skills involving degrees of mastery, as these might moderate the above relationships.

In sum, the concepts of allostasis and allostatic load very much map onto the theoretical issues that concern us.

2. What could be done to make this concept more valuable?

The first steps to improve the usefulness of the concept have already been taken in the elaboration of concepts such as primary mediators and primary, secondary, and tertiary outcomes. These help to frame what groups of variables it would be useful to study and provide the beginnings of a framework for interrelating them. In my view, these "levels" could be refined further. For example, much of the focus in secondary outcomes has been focused on measures relating to energy metabolism (e.g., insulin, lipids, and other indicators of metabolic syndrome). A second focus has been on immune dysregulation. Clearly, both classes of events are of importance in various outcomes including cardiovascular disease and possibly autoimmune diseases and cancer. Additional classes of variables deserve study. One of these includes inflammatory markers of various sorts (research on these is actually underway). A second group of mediators concerns hemostasis. Allostatic load in the hemostasis system may be indicated by increases in "procoagulatory" markers such as D-dimer and fibrinogen. Markers such as these have been linked to increased likelihood of cardiovascular disease and stroke, but have not been very systematically studied among elderly under chronic and acute stress. Other indicators of "vascular health" such as responsiveness of the endothelium to baro or hypoxic stress are also worthy of focus, since responsive endothelial functioning may protect against the sorts of injuries to vessel linings that form the precursors to arteriosclerosis.

The area of vulnerability and resilience also deserves better conceptualization. As in inherent in models of allostasis, genetic factors may predispose to chronic allostatic load in some but not others. Studies on genetics, gene expression, proteomics, and metabolomics should ultimately be done to clarify who under what circumstances responds in what manner, and how such proteomic and metabolomic changes influence downstream organ health and disease.

A great deal of work has been done in the fields of psychosomatic medicine and behavioral medicine generally on resilience factors including concepts like problem focused coping, mastery, etc. Better linkage between these notions and studies of psychobiology of aging under conditions of acute and chronic stress are needed.

3. What are the most critical measurement issues?

Research into allostasis and allostatic load would be advanced by developing normative data in regard to primary and secondary outcomes. For example, what represents "normal" response to a stressor for men and women of different ages, socioeconomic strata, educational levels, or race/ethnicity? In neuropsychology a great deal of progress has been made in the ability to detect and monitor change in neurocognitive performance in respect to illness or aging by developing expected ranges of performance on various neurocognitive tests with respect to age, education, gender, and race/ethnicity. Such standards allow us to determine whether responses of groups of interest are within the expected range, or may represent maladaptive responses. Until we have good data on how, for example, elderly men and women who are in generally good health and not under conditions of severe stress respond to acutely stressful circumstances in terms of catecholamines, glucocorticoids, ACTH, etc., it will be difficult to explore notions of change in "set point" that may be occasioned by chronic stress.

Related to this, several groups of investigators have attempted to create overall measures of allostatic load. These have involved creation of standard scores, counting of scores outside particular ranges, etc. These are useful starting points, but if we had more data on "normals" then it might be possible to generate Z or T-scores for the primary and secondary outcome measures that are more reliable.

4. What would be the most important "next steps" in moving research on allostatic load forward?

Some of these notions have been covered in #3 above. Another useful approach would be the design of experiments that are able to test the theorized mechanistic pathways linking stress to primary, secondary, and tertiary outcomes in a more precise way. From the example of hemostatic variables mentioned above, can we, for example, link changes in basal and/or stress induced hemostatic response to changes in endothelial function, vascular response, or early indicators of arteriosclerosis? Mapping these more fine-grained associations would permit moving us from studies that are correlative in nature to those in which it may be possible to test models of causality. Implied in this strategy is the importance of examining people or animal models over time with a view to accumulating sufficient longitudinal information to test such causal models.

Arun Karlamangla


The Concept

The concept of allostatic load originates from the idea that the internal physiologic milieu adapts to environmental demands – a phenomenon referred to as allostasis. Allostasis is a dynamic regulatory process, with continuous adaptation of physiology in response to stressors. However, when adaptation efforts are excessive, in terms of frequency, duration, and/or extent, it can lead to gradual loss of the body’s ability to maintain system parameters within normal operating ranges. Allostatic load is the total accumulation of such dysregulation across physiological systems, and was hypothesized to mediate the effects of stress on health risks. There is now good evidence that psychosocial adversity is associated with higher levels of dysregulation in multiple physiological systems, or higher allostatic load. Higher allostatic load (or larger number of systems that are dysregulated) is in turn, associated with poorer health outcomes.

Potential Use in Clinical Care

In addition to the research role of allostatic load in understanding the mechanisms by which psychosocial influences on health risks play out, allostatic load can also be potentially used as a measure of total sub-clinical change in physiology, to assess risk for adverse outcomes. In young adults, it could be an early warning sign of health risks accumulating beneath the surface, which should trigger changes in health behaviors. Markers of sub-clinical changes are especially important in older adults: Many older men and women experience gradual declines in physical and cognitive abilities in the absence of a clinically manifest disease process, and sub-clinical measures such as allostatic load can identify the older adults at increased risk for such declines, and target appropriate interventions.


Initial operationalizations of allostatic load have used simple count measures based on resting or baseline values of physiological parameters from multiple regulatory systems. While the approach has served to test the allostatic load concept, it likely does not adequately capture the extent of dysregulation and is insensitive to changes in system dynamics that do not affect resting values. To maximize the utility of the allostatic load concept for both research and clinical uses, the allostatic load measure has to reliably and reproducibly measure both the extent/severity of dysregulation and aspects of the dynamics in multiple physiological systems. Measurement questions that need to be answered include:

  1. How to incorporate severity of dysregulation? Should clinically recognized disease entities be treated as the most severe form?
  2. How to handle use of medications that affect the measured biomarker (since medication use might modify the relationship between an affected biomarker and health outcomes)?
  3. How to incorporate past history of dysregulation (since a longer history of dysregulation is associated with poorer health outcomes)?
  4. If some systems are better represented (by more biomarker measurements) than others, is weighting needed?
  5. How to handle interactions between physiological systems (e.g., high blood pressure has worse implications in the presence of diabetes)?
  6. How to measure reactivity to challenge? How to measure recovery? Diminished reactivity and delayed recovery are both examples of dysregulation that are not captured by an index based only on measurement of resting/baseline values.
  7. Does a high level of psychosocial well-being lead to better than normal regulation? If so, how to measure it?
  8. Validation: Should the allostatic load measure be validated w.r.t. prediction ability for health outcomes or w.r.t. dependence on psychosocial histories or both? Should mediation be the test of validation? If yes, what is the best way to test mediation?


Dario Maestripieri

University of Chicago

Interindividual variation in allostatic load:
Investigating the role of early experience with a nonhuman primate model

Understanding interindividual variation in allostatic load

Allostatic load refers to the cumulative strain on the body produced by the physiological costs of repeated adjustments to stressful perturbations as well as the elevated activity of physiological systems under challenge. Not all individuals are affected by allostatic load in the same way. "There is a growing body of correlational evidence indicating that individuals vary in their response to stressors based on differences in personality, coping and emotional regulatory styles, and social and cultural environments. Thus, some individuals seem resilient to the effects of aging and present a profile of positive health and well-being that may protect stress regulatory systems from dysregulation. Questions remain as to whether some people are genetically disposed to greater resilience to stress, or whether life-style, psychosocial, and socioeconomic factors are responsible for these differences" (Seeman & Nielsen, Background Statement for NIA Workshop on Allostatic Load).

As a developmental scientist using a lifespan approach to address questions regarding behavior, physiology, and health, I am interested in understanding the effects of early experience, in conjunction with those of genetic make-up and current environment, on the development of stress vulnerability and resilience, including the accumulation of allostatic load in aging individuals.

Early exposure to stress and the development of stress vulnerability and resilience

Exposure to stress early in life has long been known to increase vulnerability to stress and stress-related diseases later in life. Laboratory studies have repeatedly shown that the hypothalamic-pituitary-adrenal (HPA) axis is involved in mediating this link, as stressful early life events contribute to long-term changes in both brain and neuroendocrine stress responsiveness in humans and animals. It is now well established that severe stress experienced during infancy or childhood impairs the acquisition of appropriate coping skills, enhances stress-induced HPA axis activation, and increases the risk for the development of mood and anxiety disorders in adulthood, as well as other stress-related diseases.

Stress researchers have often assumed that stress vulnerability later in life increases linearly as a function of the intensity of the early stressor, so that individuals who are exposed to severe traumatic events will exhibit later greater vulnerability than individuals exposed to mild stressors, who in turn will be more vulnerable than individuals exposed to little or no stress (Fig. 1a). Several studies, however, have shown that prior stress exposure does not increase vulnerability in a linear fashion, but according to a J-shaped curvilinear function (Fig. 1b). Whereas severe early life stress exposure generally undermines the development of resilience and leads to vulnerability, mild early life stress exposure may protect against these deleterious effects. Specifically, milder forms of adversity may provide a challenge, that when overcome, produces competence in the management of, and enhanced resistance to, subsequent stressors.

The idea that mild early life stress exposure may "inoculate" the developing organism by permanently altering cognitive appraisal of, and emotional and neuroendocrine sensitivity to subsequent stressors has potentially profound implications for understanding the accumulation of allostatic load and the occurrence of stress-related diseases late in life. Early life stress inoculation may especially confer psychological health benefits for women, who more often engage in maladaptive ruminative coping strategies that increase negative emotional arousal and exhibit greater HPA axis responses to pharmacological stimulation and social rejection than men. Moreover, after puberty, women are twice as likely as men to develop certain stress-related mood and anxiety disorders.

As a comparative behavioral scientist working with animal models of human behavior, development, and health, I am interested in developing a nonhuman primate model to understand how early exposure to stressors of different intensity affects the development of stress vulnerability and resilience and the accumulation of allostatic load in aging individuals, particularly females.

A nonhuman primate model of early experience and allostatic load

My model organism is the rhesus macaque and my model for studying allostatic load is dominance rank, the equivalent of socioeconomic status in humans. In rhesus macaques, baboons, and related species, female dominance rank is acquired early in life and remains stable throughout the lifespan. In these species, chronic stress associated with low dominance rank has a number of adverse adrenocortical, cardiovascular, reproductive, immunological, and neurobiological consequences. Low ranking monkeys, particularly females, exhibit basal hyperactivity of the HPA axis and cathecolaminergic systems, basal hypertension and elevated heart rate, blunted cardiovascular stress response after a challenge, pathogenic cholesterol profiles, increased vulnerability to the atherogenic effects of a high fat diet, later timing of puberty and decreased gonadal hormone levels in adulthood, suppression of circulating lymphocyte numbers and blunted immune responsiveness to a challenge, and a host of neurobiological changes including inhibition of neurogenesis, dendritic atrophy, and impairment of synaptic plasticity in the hippocampus. The deleterious physiological consequences of chronic stress associated with low dominance rank should be particularly evident in aging individuals, although few or no studies so far have investigated allostatic load and aging in free-ranging primate populations.

I am currently involved in two research projects investigating early experience and allostatic load in aging monkey females at the Caribbean Primate Research Center in Puerto Rico. My subject population is the free-ranging rhesus macaque population (n= 850) on the island of Cayo Santiago, which has been the subject of behavioral and biomedical research since the 1960s.

The first project focuses on aging females (all of the oldest females in the population) and involves the collection of multiple biomarkers of allostatic load in these individuals. We expect that aging females of low dominance rank will exhibit greater evidence of allostatic load than same-aged females of high dominance rank, as well as than younger females of both low and high dominance rank. We also expect that there will be significant inter-individual variation in biomarkers of allostatic load in all aging females, but especially in those of low rank. We measure basal levels of glucocorticoid hormones in fecal samples, and plasma HPA axis hormone concentrations in response to various challenges. Peptide and monoamine metabolite concentrations in the CSF are measured. Proinflammatory cytokines and other biomarkers of health are measured in blood and CSF samples.

The second project is a prospective longitudinal study aimed at investigating whether exposure to psychosocial stressors of different intensity in infancy can account for a significant fraction of interindividual variation in allostatic load among aging females of low dominance rank. This project involves 2 phases: in the first phase, we investigate whether exposure to stressors of different intensity in infancy affects the development of behavioral and physiological measures of stress responsiveness in the first 3 years of life (i.e. birth through puberty). In a subsequent phase of the project, study subjects will be followed into old age (15-25 years) to examine whether early exposure to stress and individual profiles of responsiveness to stress in the first 3 years of life predict differential accumulation of allostatic load in old age, especially among females of low rank.

Two contrasting models of the relation between early stress and later vulnerability/resilience will be tested: one in which there is a linear relationship between intensity of early stress and later vulnerability (Fig. 1a) and another in which this relationship is a J-shaped curvilinear function (Fig. 1b). A crucial prediction of the second model is that exposure to mild stress early in life will promote later resilience and function as a protective factor against the accumulation of allostatic load in females of low rank. Other risk/protective factors being considered are genotype (the polymorphisms in the serotonin transporter gene) and current environment (the availability or lack of social support).

We use a naturalistic model of variable early stress exposure: naturally occurring interindividual variation in maternal rejection behavior. Mothers prevent their infants from making contact with them and nursing by pushing them away, and occasionally hitting and biting them. Rejected infants exhibit intense distress behavior including prolonged screaming bouts and tantrums. There is great interindividual variability in maternal rejection behavior. Although the general function of maternal rejection is to encourage infant independence, some mothers exhibit extremely high rates of rejection, often in conjunction with physical abuse of their infants. Previous studies have shown that exposure to variable maternal rejection in the first few months of life affects behavioral reactivity to novelty and stress later in life, the development of the HPA axis, and the development of the brain serotonergic and dopaminergic system. Infants exposed to mild rates of maternal rejection seem to benefit in terms of greater independence from their mothers, while those who experience high rates of aggression exhibit hyper-reactivity to stress, fear and anxiety, and associated neuroendocrine and neurobiological alterations. The deleterious consequences of early exposure to high maternal rejection appear to be greater in individuals carrying the short allele of the serotonin transporter gene, and in individuals lacking large and effective networks of social support.

Jan Moynihan

University of Rochester Medical Center

Until relatively recently, our lab has focused on animal psychoneuroimmunology, using rodent models to dissect out relationships among stress, the central nervous system, and immune responses. The use of mouse models has in general been a fruitful basic science approach with important translational implications. Our science was originally informed by the General-Adaptation-Syndrome theory of Hans Selye 1; to a large extent, the concept of allostatic load can be argued to be the updated version of this theory.
To summarize my own findings and those of others over the years, the effects of stress over time (arguably synonymous with cumulative wear and tear resulting in ‘allostatic load’) on immunity in genetically inbred mice (as well as ‘outbred’ humans) are dependent upon at least three sets of factors. The first concerns the experimental subject, and includes: species and genetic strain; age and gender; previous life history; and circadian rhythm (the time of day at which a stressor is imposed). A second set of variables has to do with the stressor itself, including: the intensity and duration of the stressor (acute versus chronic); the timing of stressor in relation to (immunological) challenge; and, perhaps most importantly, the subject’s perception of, and capacity to cope with, the stressor. Finally, the immunological (or other physiological) measure of interest itself presents another important set of factors in determining the outcome of stressful encounters, including: the kinetics of the immune response, the nature of the response (i.e., innate versus cell-mediated versus humoral); the antigen used to elicit a response and its concentration; the site of response (peripheral blood versus lymphoid organ); and the use of in vitro versus in vivo measures or ex vivo measurements. The concept of allostatic load provides us with a theoretical framework in which to think about the dynamic nature of these interacting factors.

Neuroendocrine responses to stressful stimuli may vary as a function of the stressor. The concept of allostatic load—cumulative wear and tear on physiological systems and organs-- raises the question of how environmental stimuli are translated into physiological or immunological changes. We understand that there are two important CNS-derived responses to stressors: activation of the hypothalamo-pituitary-adrenal (HPA) axis, and activation of the autonomic nervous system. With respect to immune function, activation of these neurochemical pathways, and the release of virtually every one of their hormones and transmitters, has the potential to alter some aspect of immune response 2;3. Of all these neurochemicals, adrenally-derived corticosteroids (corticosterone in rodents and cortisol in humans) have received overwhelming attention from investigators searching for the mechanism underlying stress-mediated immune alteration. Certainly, there are data to support the hypothesis that increased corticosterone levels, either alone or in concert with other hormones, mediate altered immune function 4-6. However, we should not forget older papers in the literature that have convincingly shown that stress-induced changes in immunity can occur in either adrenalectomized or hypophysectomized animals (e.g., 7-9). Further, stress-induced elevations in glucocorticoid levels can occur in the absence of detectable changes in immune function 10;11. Finally, stress-induced immune alteration can occur in the absence of a detectable, or perhaps biologically meaningful, change in glucocorticoid levels 12-14. Thus, experimental designs that focus on a single neurochemical mediator fall short of providing a complete understanding of allostatic load. Further, pharmacotherapies that might target a single "stress hormone" will be unlikely to rescue immune or other physiological functions that are impaired by allostatic load.
Another important question that arises from thinking about allostatic load is: what is the nature of the relationship between an experimental subject and an individual stressor as a function of time or aging? There are some stressors that an experimental subject is likely to repeatedly perceive as stressful (for scientists, a good example may be the process of NIH grant submission). In contrast, there are some stressors to which a subject will learn to adapt or habituate. Finally, there are some stressors to which a subject learns maladaptive behaviors (e.g., learned helplessness). How are these disparate behaviors related to, or perhaps driven by, the neuroendocrine response to that stressor?
Escapable versus inescapable shock is a paradigm involving pairs of rats, one of which learns to turn off electric shock (also called the "executive animal"); the other, yoked control rat is not allowed to learn, receiving an identical amount of shock that is inescapable. The paradigm provides an interesting model for examining the neuroendocrine mechanism(s) underlying a) the ability to control or cope with a stressor (escapable stress), versus b) the induction of learned helplessness (the subsequent inability to terminate or escape from the stressor when given the opportunity to do so). Rodents subjected to escapable versus inescapable shock differ in performance of a number of subsequent behavioral tasks; as one example, yoked animals show a decrease in social interaction compared to executive animals 15-17. Despite these clear behavioral differences, however, the HPA axis response of these two very different rats to the shock experience is similar; that is, the magnitude and duration of the adrenocorticotropic hormone (ACTH) and corticosterone response to shock does not discriminate between the two groups 18;19.

In contrast to similar HPA axis hormone levels observed in executive and yoked animals, Dr. Dana Helmreich (University of Rochester Center for Mind-Body Research (RCMBR)) and her colleagues have observed differential regulation of another relatively unstudied stress-responsive pathway, the hypothalamo-pituitary-thyroid (HPT) axis, in rats subjected to escapable or inescapable shock. That is, peripheral T3 levels, which are critical for maintenance of homeostasis throughout the lifespan, do differ between executive and yoked animals 20; T3 levels are decreased in rats subjected to inescapable shock. Dr. Helmreich’s results are similar to those reported by Josko 21; yoked rats have a decrease in circulating thyroid stimulating hormone (TSH), T3 and T4 levels compared to executive rats.

In humans, dysregulation of thyroid hormones is associated with mood disorders, particularly depression 22. Recently, modest decreases in thyroid hormones, yet still within clinical boundaries of euthyroidism, have been correlated with increased incidence of post-partum depression 23, increased anxiety 24, decreased well-being and quality of life 25;26 and increases in metabolic syndrome 27. These results support the conclusion that subtle changes in thyroid hormone levels have quantifiable and significant consequences for mental health, and perhaps immunological health as well.
These data suggest that different neuroendocrine pathways may be activated by various stressful stimuli. These patterns of activation are undoubtedly associated with differing patterns of, for example, immune outcome measures. Thus, rather than examining a single biomarker or immune outcome, our understanding of stress-induced changes in health and disease (the consequence of allostatic load) will require the examination of a combination of neuroendocrine and immunological outcome measures.

The role of genetics in the response to a stressor(s). Even a simple manipulation such as the differential housing of genetically inbred strains of mice can illustrate the complex interactions among environment, genetics and immune function. In one such experiment, we housed male BALB/c and C57Bl/6 mice 1/cage versus 4/cage. Their secondary antibody response following immunization with a novel protein antigen was assayed. BALB/c mice responded with a robust antibody response; no differences were observed between BALB/c mice housed 1/cage versus 4/cage. C57Bl/6 mice responded to the same concentration of antigen with a significantly lower antibody response than BALB/c mice, regardless of housing condition. More importantly, C57Bl/6 mice housed 1/cage produced a significantly higher antibody response than C57Bl/6 mice housed 4/cage at all time points examined 28.

The direction of the difference in antibody titers between singly versus group-housed C57Bl/6 mice may have seemed surprising to us initially; perhaps we might have expected that mice housed alone would be "isolated," and should therefore have lower, not higher, antibody levels. The lack of an effect of housing for the BALB/c mice becomes more interesting in light of a recent paper which suggests that compared to C57Bl/6 mice, BALB/c mice are low in sociability 29. One might speculate that for the BALB/c mouse, being housed 1/cage versus 4/cage is much less "disturbing" than it is to the C57Bl/6, social animal.

The role of epigenetics in responses to stressors. A large body of work from Michael Meaney and his colleagues documents that early experiences, including the early rearing environment, can cause changes in gene expression in stress-sensitive brain regions in rats that last into adulthood 30-32. The implications of these changes in epigenetic programming for neuroendocrine, and especially immune responses and health, remain largely unknown. As the perinatal period is a period of plasticity for both the nervous and immune systems, it is not difficult to imagine that early events might also directly affect immunity, and/or that altered central nervous system function, e.g., altered expression of glucocorticoid receptors in the brains of adult rodents, would influence immunity across the life course.

Clinical studies that examine early events, in this instance during pregnancy, are ongoing in our Laboratory for Mental Disorders at the University of Rochester Medical Center, under the direction of Thomas O’Connor, PhD. Dr. O’Connor has shown an association between maternal stress and anxiety during pregnancy with long-term cortisol dysregulation and behavior problems in offspring, who are now 10-12 years of age 33;34. Dr. O’Connor extends these findings in a new study, Prenatal Anxiety and Its Effects on Child Development, which will examine the effects of anxiety during pregnancy on child outcomes in infancy, including the immune response to hepatitis B vaccination.
In addition to these human studies, my colleagues in Rochester are conducting rodent experiments to determine (1) how early life experiences influence developmental and long-term behavioral, neuroendocrine and immunological processes, and how changes in function might be influenced, positively and negatively, by subsequent experiences to alter the development and/or progression of disease throughout the life span; and (2) the developmental and long-term psychophysiological and health effects of immunological challenges experienced during early life. This work is conducted by my colleague in the RCMBR, David Parfitt, PhD, who is particularly interested in neonatal rearing and adult stress reactivity in C57BL/6 mice. We are proposing longitudinal studies in rodent models that test hypotheses about early life events across the lifespan in ways that cannot be studied in humans.

Conclusions. One approach to understanding allostatic load and aging utilizes translational studies in animals that inform the process of "wear and tear" and aging in humans. While the complexity of interacting variables—not to mention the redundancy that is built into many of our systems (such as is exhibited in the proinflammatory cytokine cascade)—may appear daunting, it is important to understand how a lifetime burden of stressors impacts the aging process. In tandem, studies should consider ameliorating these potentially deleterious effects, either pharmacotherapeutically or via psychological or physical (e.g., exercise) interventions.

Reference List

1. Selye H. Stress. Montreal: ACTA, 1950.
2. Levite M. Neuropeptides, by direct interaction with T cells, induce cytokine secretion and break the commitment to a distinct helper phenotype. Proc.Natl.Acad.Sci. 1998;95:12544-49.
3. Khansari DN, Murgo AJ, Faith RE. Effects of stress on the immune system. Immunol.Today 1990;11:170-75.
4. Dobbs CM, Feng N, Beck M, Sheridan JF. Neuroendocrine regulation of cytokine production during experimental influenza viral infection: effects of restraint stress-induced elevation in endogenous corticosterone. J.Immunol. 1996;157:1870-77.
5. Coe CL, Rosenberg LT, Fisher M, Levine S. Psychological factors capable of preventing the inhibition of antibody responses in separated infant monkeys. Child.Dev. 1987;58:1420-30.
6. Blecha F, Kelley KW, Satterlee DG. Adrenal involvement in the expression of delayed-type hypersensitivity to DNFB in mice. Proc.Exp.Biol.Med. 1982;169:247-52.
7. Keller SE, Weiss JM, Schleifer SJ, Miller NE, Stein M. Stress-induced immunosuppression of immunity in adrenalectomized rats. Science 1983;221:1301-04.
8. Esterling B, Rabin BS. Stress-induced alteration of T-lymphocyte subsets and humoral immunity in mice. Behav.Neurosci. 1987;101:115-19.
9. Rabin BS, Cunnick JE, Lysle DT. Stress-induced alteration of immune function. NeuroendocrinImmunol. 1990;3:116-24.
10. Mormede P, Dantzer R, Michaud B, Kelley K, LeMoal M. Influence of stressor predictability and behavioral control on lymphocyte reactivity, antibody responses, and neuroendocrine activation in rats. Physiol.Behav. 1988;43:577-83.
11. Flores CM, Hernandez MD, Hargreaves KM, Bayer BM. Restraint stress-induced elevations in plasma corticosterone and beta-endorphin are not accompanied by alterations in immune function. J.Neuroimmunol. 1990;28:219-25.
12. Jessop JJ, Gale K, Bayer BM. Enhancement of rat lymphocyte proliferation after prolonged exposure to stress. J.Neuroimmunol. 1987;16:261-71.
13. Moynihan JA, Koota D, Brenner GJ, Cohen N, Ader R. Repeated intraperitoneal injections of saline attenuate the antibody response to a subsequent intraperitoneal injection of antigen. Brain Behav.Immun. 1989;3:90-96.
14. Moynihan JA, Karp JD, Cohen N, Ader R. Immune deviation following stress odor exposure: role of endogenous opioids. J.Neuroimmunol. 2000;102:145-53.
15. Haller J, Bakos N. Stress-induced social avoidance: a new model of stress-induced anxiety? Physiol Behav. 2002;77:327-32.
16. Blatchford KE, Diamond K, Westbrook RF, McNally GP. Increased vulnerability to stress following opiate exposures: behavioral and autonomic correlates. Behav.Neurosci. 2005;119:1034-41.
17. Short KR, Maier SF. Stressor controllability, social interaction, and benzodiazepine systems. Pharmacol Biochem.Behav. 1993;45:827-35.
18. Prince CR, Anisman H. Situation specific effects of stressor controllability on plasma corticosterone changes in mice. Pharmacol Biochem.Behav. 1990;37:613-21.
19. Maier SF, Ryan SM, Barksdale CM, Kalin NH. Stressor controllability and the pituitary-adrenal system. Behav.Neurosci. 1986;100:669-74.
20. Helmreich DL, Crouch M, Dorr NP, Parfitt DB. Peripheral triiodothyronine (T(3)) levels during escapable and inescapable footshock. Physiol Behav. 2006;87:114-19.
21. Josko J. Liberation of thyreotropin, thyroxine and triiodothyronine in the controllable and uncontrollable stress and after administration of naloxone in rats. J Physiol Pharmacol 1996;47:303-10.
22. Joffe RT, Sokolov ST. Thyroid hormone treatment of primary unipolar depression: a review. Int J Neuropsychopharmacol. 2000;3:143-47.
23. Pedersen CA, Johnson JL, Silva S, Bunevicius R, Meltzer-Brody S, Hamer RM et al. Antenatal thyroid correlates of postpartum depression. Psychoneuroendocrinology 2007;32:235-45.
24. Sait GM, Kisakol G, Savas CA, Dikbas O, Gungor K, Inal A et al. Assessment of anxiety in subclinical thyroid disorders. Endocr.J 2004;51:311-15.
25. Gulseren S, Gulseren L, Hekimsoy Z, Cetinay P, Ozen C, Tokatlioglu B. Depression, anxiety, health-related quality of life, and disability in patients with overt and subclinical thyroid dysfunction. Arch.Med Res 2006;37:133-39.
26. Saravanan P, Visser TJ, Dayan CM. Psychological well-being correlates with free thyroxine but not free 3,5,3'-triiodothyronine levels in patients on thyroid hormone replacement. J Clin Endocrinol.Metab 2006;91:3389-93.
27. Lin SY, Wang YY, Liu PH, Lai WA, Sheu WH. Lower serum free thyroxine levels are associated with metabolic syndrome in a Chinese population. Metabolism 2005;54:1524-28.
28. Karp JD, Moynihan JA, Ader R. Effects of differential housing on the primary and secondary antibody responses of male C57BL/6 and BALB/c mice. Brain Behav.Immun. 1993;7:326-33.
29. Brodkin ES. BALB/c mice: low sociability and other phenotypes that may be relevant to autism. Behav.Brain Res 2007;176:53-65.
30. Weaver IC, Cervoni N, Champagne FA, D'Alessio AC, Sharma S, Seckl JR et al. Epigenetic programming by maternal behavior. Nat.Neurosci. 2004;7:847-54.
31. Weaver IC, Champagne FA, Brown SE, Dymov S, Sharma S, Meaney MJ et al. Reversal of maternal programming of stress responses in adult offspring through methyl supplementation: altering epigenetic marking later in life. J Neurosci. 2005;25:11045-54.
32. Szyf M, Weaver IC, Champagne FA, Diorio J, Meaney MJ. Maternal programming of steroid receptor expression and phenotype through DNA methylation in the rat. Front Neuroendocrinol. 2005;26:139-62.
33. O'Connor TG, Heron J, Golding J, Glover V. Maternal antenatal anxiety and behavioural/emotional problems in children: a test of a programming hypothesis. J Child Psychol Psychiatry 2003;44:1025-36.
34. O'Connor TG, Ben Shlomo Y, Heron J, Golding J, Adams D, Glover V. Prenatal anxiety predicts individual differences in cortisol in pre-adolescent children. Biol.Psychiatry 2005;58:211-17.

Julian Thayer

The Ohio State University

The Allostatic Load Model has been important in focusing researchers on the concept of cumulative physiological risk. Of special note is its attention to the temporal dimension. In our own work we have proposed the Prolonged Activation Model with an emphasis on anticipatory stress and delayed recovery, in the context of our broader Neurovisceral Integration Model. Similar to the Allostatic Load Model, our Prolonged Activation Model seeks to identify the pathogenic state that leads to organic disease. We feel that the cumulative load put on the system is a critical factor on the path from psychological stress to morbidity and mortality.

There are several areas where our models have focused that may be under-represented in the Allostatic Load Model. One area is individual differences. Whereas the Allostatic Load Model uses individual differences in allostatic load scores to predict morbidity and mortality, we use individual differences in heart rate variability (HRV), for example, to predict differences in stress reactivity per se including anticipatory stress and delayed recovery. Thus we are able to look at individual differences in the regulation of allostatic systems at a more fine grained level than is common in the Allostatic Load Model. This has implications for the understanding of gender and ethnic differences in the accumulation of risk.

A second area is the relationship between the central nervous system and the peripheral nervous system. Our studies of simultaneous measurement of cerebral blood flow or other indices of CNS function and peripheral measures such as HRV allow us to make inferences about the central concomitants of the regulatory processes being indexed by our peripheral measures. We are also able to show how these central and peripheral interactions might be changing with aging or as a function of genotype.

A third area involves the cognitive processes that lead to prolonged activation. In this context our Perseverative Cognition Hypothesis provides insights into the toxic cognitive mechanism that leads to prolonged stress responses and thus the excess wear and tear (cumulative load) on the system. What is actually going on inside the heads, as it were, and how to modify those thought processes represent potential avenues of intervention to decrease cumulative risk.

A fourth area is the use of nonlinear dynamical systems theory and especially the role of inhibitory processes. The interplay of excitatory and inhibitory processes is essential for the proper functioning of a nonlinear dynamical system. Importantly, a little bit of inhibition at the right place and/or the right time can be a critical factor in stress responsivity and whether it leads to prolonged activation or not. Nonlinear dynamical systems theory also provides a broader perspective for the generation of testable hypotheses about the pathways and mechanisms associated with cumulative physiological risk. From a measurement perspective the integration of indices of the parasympathetic nervous system should enhance the measurement of allostatic load.

A fifth area involves possible interventions that could help to reduce cumulative risk. We are currently testing both psychological (e.g., worry reduction and CBT) as well as behavioral (dietary e.g., omega-3 and exercise) interventions to decrease markers of cumulative risk. Interventions aimed at improving allostatic load scores might be fruitfully employed to reduce cumulative risk as well.

And finally, the next steps for moving research on cumulative risk forward include funding mechanisms that allow for such interdisciplinary research to take place. The current disease specific funding mechanisms can not accommodate the type of integrative models that the allostatic load model exemplifies. Further incorporation of these ideas into the allostatic load concept could make it more valuable to researchers.

The Social Environment and Biomarkers of Aging Study (SEBAS) is collaborative with the Bureau of Health Promotion in Taichung, Taiwan and includes Maxine Weinstein, Dana Glei, and myself as leading investigators.

close window


UCSF Home About UCSF Search UCSF UCSF Medical Center