Original article
Allostatic load as a predictor of functional decline: MacArthur studies of successful aging

https://doi.org/10.1016/S0895-4356(02)00399-2Get rights and content

Abstract

Allostatic load has been proposed as a cumulative measure of dysregulation across multiple physiological systems, and has been postulated to impact health risks. In the allostatic load model, increased risk is hypothesized to result not only from large and clinically significant dysregulation in individual systems, but also from more modest dysregulation, if present in multiple systems. Our objective was to construct an allostatic load score by optimally combining several physiologic measurements, and to examine its association with future functional decline. We analyzed data from a 7-year longitudinal study of a community-based cohort, whose age at baseline was between 70 and 79 years. Canonical correlation analysis was used to study the association of 10 biological measurements representing allostatic load with declines in scores on five tests each of physical and cognitive function over two follow-up periods: 1998–1991 and 1991–1995. We used bootstrapping to evaluate the stability of the canonical correlation and canonical weights. The canonical correlation between allostatic load and the 20 decline scores was 0.43 (P = .03) and the [25th, 75th] percentile interval of its distribution over 200 bootstrapped subsamples of the cohort was [0.48, 0.53]. These findings were not substantially affected by adjusting for covariates and cardiovascular disease. We conclude that a summary measure of physiologic dysregulation, such as allostatic load, is an independent predictor of functional decline in elderly men and women.

Introduction

With the aging of the world's population and the emerging patterns of increasing dysregulation with age in several physiological systems 1, 2, 3, there has been growing interest in the role of physiologic dysregulation in age-associated declines in health status and functioning level. While dysregulated blood pressure and glucose and lipid metabolism are well-known risk factors for the incidence and consequences of cardiovascular disease, dysregulation in hypothalamic–pituitary–adrenal (HPA) axis activity has also been found to be associated with cardiovascular disease incidence [4], cognitive decline [5], and physical frailty [6]. Similarly, chronic sympathetic activation is believed to contribute to the incidence of coronary artery disease [7] and susceptibility to infectious disease [8]. Because a comprehensive measure that incorporates multiple risk factors is expected to better predict future health risks than any single risk factor by itself [9], allostatic load was proposed by McEwen and Stellar as a cumulative measure of physiologic dysregulation across multiple systems, and was hypothesized to have considerable impact on future health risks [10].

The idea of allostatic load is derived from the concept of allostasis introduced by Sterling and Eyer in 1988, as the body's ability to adapt its internal physiologic milieu to match external demand [11]. This is an extension of the concept of homeostasis or constancy of the physiologic milieu. Many physiologic parameters (such as blood pressure) do not remain constant, and instead, vary appreciably in response to perceived challenges. The normal physiologic response to a challenging situation is after all, a state of arousal, as in the fight or flight response. Allostasis is this dynamic regulatory process, with continuous evaluation of need and adaptation of physiologic set points.

Allostatic adaptation of the body to repeated challenges is, however, not without cost. When adaptation efforts are excessive (either in terms of frequency, duration, or extent) it can lead to gradual loss of the body's ability to maintain system parameters within normal operating ranges, with respect to both resting levels and dynamic patterns of response to stressors. For instance, when challenges are frequent or long-standing, the response can gradually become exaggerated or muted, its turn-off can become inefficient, or the resting level itself may get altered. Frequent or chronic arousal has been found to be associated with ultimate dysregulation of several major physiological systems, including the HPA axis [12], the sympathetic nervous system 13, 14, and the immune system [15]. The concept of allostatic load was introduced as an attempt to view the cumulative impact of health risks from such dysregulation in multiple physiological systems.

The physiologic agents that mediate arousal have protective and adaptive effects in the short term, but can accelerate pathophysiology over the long term, if they are produced insufficiently or in excess, i.e., outside of their normal ranges. In the allostatic load formulation, impact on health risk is thought to result not only from large and clinically significant deviations from normal operating ranges, but also from more modest dysregulation, if it is present in multiple systems. McEwen and Stellar hypothesized that the cumulative impact on health risk from modest dysregulations in multiple systems can be substantial, even if they individually have minimal and insignificant health effects. Accordingly, they defined allostatic load as a cumulative measure of physiologic dysregulation over multiple systems.

The idea that allostatic load accumulates over time was derived from the frequently observed patterns of growing dysregulation in multiple physiological systems at later ages 16, 17. Examples of this include age-related declines in heart rate (and blood pressure) variability, increases in glucose intolerance [18], fasting blood glucose levels, and blood glycosylated hemoglobin levels [19], and altered pulsatility of hormones such as growth hormones. Also, plasma norepinephrine levels and sympathetic tone [20] tend to rise with age, while serum levels of dehydroepiandosterone sulfate (DHEA-S) tend to fall by 10% per decade of life [21]. After a perturbation, the HPA axis tends to take longer to return to baseline in older men and women compared to younger individuals [22], just as postprandial blood glucose levels tend to take longer to return to fasting levels in older men and women [18].

Previously, we introduced an initial summary measure of allostatic load based on 10 markers that reflect alterations in levels of activity of physiological systems that have individually been linked to increased risks for disease [23]. These include: (1) 12-hr overnight urinary cortisol excretion, (2) 12-hr overnight urinary excretion of norepinephrine, (3) 12-hr overnight urinary excretion of epinephrine, (4) serum DHEA-S, (5) average systolic blood pressure, (6) average diastolic blood pressure, (7) ratio of waist-to-hip circumference, (8) serum high density lipid (HDL) cholesterol, (9) ratio of total-to-HDL cholesterol in the serum, and (10) blood glycosylated hemoglobin. This list is necessarily incomplete as it is a reflection of the measurements that were available for assessing allostatic load in a secondary data analysis study [23]. With respect to the measurements included in this initial index of allostatic load, cortisol levels for instance, are known to increase when an individual is under stress: an adaptive response intended to replinish energy stores after fight or flight. However, if cortisol is chronically elevated, it impedes insulin and promotes obesity [24], hypertension, diabetes mellitus type 2, lipid imbalance, and atherosclerosis 25, 26, 27. Chronic elevations of cortisol levels also lead to brain aging [28], hippocampal atrophy and cognitive impairment 29, 30, 31, loss of bone mineral density [32], sarcopenia [33], and immune dysfunction [34]. Similarly, catecholamine levels increase under stress, but if chronically elevated they can raise blood pressure and heart rate, which promotes atherosclerosis and appears to be independently associated with cancer mortality [35]. Dehydroepiandosterone and its sulfate ester DHEA-S are functional antagonists of cortisol 36, 37. Chronically low levels of DHEA-S are thought to be deleterious [38], and have been associated with cardiovascular disease and mortality 39, 40. The other six variables in the above list (systolic and diastolic blood pressure, glycosylated hemoglobin, waist–hip ratio, HDL cholesterol, and total-to-HDL cholesterol ratio) are markers of dysregulation in other physiologic systems, and are traditional risk factors for cardiovascular disease. A detailed exposition of the biological pathways underlying the allostatic load model, including the putative local mediators (cytokines, amino acids, and neuropeptides) has been previously published [41].

The summary measure of allostatic load used was the number (between 0 and 10) of the above 10 measurements for which the individual is in the highest risk quartile (i.e., lowest quartile for serum HDL cholesterol level and serum DHEA-S level and highest quartile for the other eight variables). The quartile cutoff points were determined from the distribution of the variables among relatively high-functioning men and women, aged 70 to 79 years, collected between 1988 and 1989, as part of the MacArthur Studies of Successful Aging. Individually, these cutoff criteria detect modest deviations from normal, but taken together, they are expected to be indicative of a cumulative level of dysregulation that will be predictive of poor health over the long term.

Indeed, while none of the ten components of allostatic load exhibited significant associations on their own with health outcomes, the summary measure of allostatic load was found to be significantly associated with four major health outcomes: (1) new cardiovascular events, (2) decline in cognitive functioning, (3) decline in physical functioning, and (4) mortality over both 2.5 and 7-year follow-ups 23, 42. These findings are consistent with the idea that although a modest deviation in the level of activity of a single physiologic system may not be predictive of poor future health, the cumulative toll from modest alterations in several physiologic systems, is indeed prognostic of poor health.

The objectives of the following analyses were to further delineate the association between allostatic load and changes in physical and cognitive functioning, and to examine an alternate summary allostatic load score that could more effectively capture this association. In particular, we address the following three questions.

Does the magnitude of the dysregulation in individual systems have predictive ability for future health? The summary measure of allostatic load used in our earlier analyses did not incorporate information about the magnitude of the dysregulation within an individual system. Each physiologic measurement contributed either a 1 or a 0 to the summary allostatic load score, depending on whether or not the measurement was in its highest risk quartile. However, in the allostatic load model, even small dysregulations are expected to contribute to health risks, if they are present in multiple systems. Hence, we propose to construct the summary allostatic load score as a linear combination of the physiologic measurements, to reflect the magnitude of dysregulation (both small and large) within individual systems.

Do different physiological systems contribute differentially to the relation between allostatic load and health outcomes? Although the equally weighted measure used in our earlier analyses was found to predict subsequent health outcomes, it is not clear that it is the best way to combine the contributions of the 10 allostatic load components to health risks. It is possible that some systems are more critical than others with respect to certain health outcomes. Here, we use canonical correlation analysis to allow for nonuniform weights for the allostatic load components. A further value of canonical analyses is that it also permits us to examine relationships between allostatic load and a weighted combination of outcome measures that reflect changes in the level of functioning in different physical and cognitive domains. Canonical correlation analyses has been used previously to study associations between combinations of psychosocial stress variables and combinations of health outcomes [43].

Do components of allostatic load, other than traditional cardiovascular risk factors, contribute to prediction of health risks? Because 6 of the 10 components of allostatic load (namely average systolic and diastolic blood pressure, waist–hip ratio, serum HDL, total cholesterol-to-HDL ratio, and blood glycosylated hemoglobin) are characteristic of Syndrome X [44] (also referred to as the metabolic syndrome [45] and the insulin resistance syndrome [46]) and represent standard risk factors for cardiovascular disease, we also examine the question of whether the other four components of allostatic load (namely serum DHEA-S and 12-hr urinary cortisol, norepinephrine, and epinephrine excretion) contribute independently to the prediction of health outcomes.

Section snippets

The study sample

Data for these analyses come from the MacArthur Study of Successful Aging, a longitudinal study of relatively high-functioning women and men, aged 70–79 years. Detailed descriptions of the study are available elsewhere [47]. Briefly, more than 4,000 men and women in this age group from three communities in the Eastern United States (Durham, NC, East Boston, MA, and New Haven, CT) were screened on the basis of four criteria of physical functioning and two criteria of cognitive functioning, to

Descriptive statistics

Data at baseline (1988/1989) for all 10 components of allostatic load were available for 729 of the 1,189 individuals in the MacArthur cohort of high-functioning men and women. Analyses comparing these individuals to the complete cohort suggest that they are generally representative of the full cohort [23]. The sample of 729 was 49% male, 81% White, and had mean education of 10.7 years (range 0–17 years, standard deviation 3.3 years). Forty percent of them reported annual income less than $10K,

Discussion

In these analyses, our objectives were: (1) to identify combinations of biological markers of dysregulation (allostatic load components) that are most predictive of functional decline, and (2) to investigate whether a multisystem measure of physiologic dysregulation, such as allostatic load, contains more prognostic information regarding functional outcomes, than is already available from standard cardiovascular risk factors alone.

Using canonical correlation analyses, we found substantially

Acknowledgements

Work on this article was supported by NIH/NIA Mentored Clinical Scientist Development Award 1K12AG01004, NIA Grants AG-17056 and AG-17265, and the MacArthur Research Networks on Successful Aging and on SES and Health through grants from the John D. and Catherine T. MacArthur Foundation.

References (75)

  • L.F Berkman et al.

    High, usual, and impaired functioning in community-dwelling older men and womenfindings from the MacArthur Foundation Research Network on Succesful Aging

    J Clin Epidemiol

    (1993)
  • M.E Williams et al.

    Manual ability as a marker of dependency in geriatric women

    J Chron Dis

    (1982)
  • H.L Taylor et al.

    A questionnaire for the assessment of leisure-time physical activity

    J Chronic Dis

    (1978)
  • H.M Perry

    The endocrinology of aging

    Clin Chem

    (1999)
  • L.A Lipsitz et al.

    Loss of “complexity” and agingpotential applications of fractals and chaos theory to senescence

    JAMA

    (1992)
  • D.A Gruenewald et al.

    Aging of the endocrine system

  • J.R Kaplan et al.

    Role of sympathadrenal medullary activation in the initiation and progression of atherosclerosis

    Circulation

    (1991)
  • S Cohen et al.

    Health psychologypsychological factors and physical disease from the perspective of human psychoneuroimmunology

    Annu Rev Psychol

    (1996)
  • B.S McEwen et al.

    Stress and the individualmechanisms leading to disease

    Arch Intern Med

    (1993)
  • P Sterling et al.

    Allostasisa new paradigm to explain arousal pathology

  • M Ockenfels et al.

    The effect of chronic stress associated with unemployment on salivary cortisoloverall cortisol levels, diurnal rhythm, and acute stress reactivity

    Psychosom Med

    (1995)
  • S.B Manuck et al.

    The pathogenicity of behavior and its neuroendocrine mediationan example from coronary artery disease

    Psychosom Med

    (1995)
  • U Lundberg

    Urban commutingcrowdedness and catecholamine secretion

    J Hum Stress

    (1976)
  • J.W Mason

    Organization of endocrine mechanisms

    Psychosom Med

    (1968)
  • J.R Rowe et al.

    Sympathetic nervous system and aging in man

    Endocr Rev

    (1980)
  • T.E Seeman et al.

    Aging and hypothalamic–pituitary–adrenal response to challenge in man

    Endocr Rev

    (1994)
  • D Elahi et al.

    Effect of age and obesity on fasting levels of glucose, insulin, glucagon, and growth hormone

    J Gerontol

    (1982)
  • B.B Arnetz

    The influence of aging on hemoglobin A1C levels

    J Gerontol

    (1982)
  • J.P Hinson et al.

    DHEA deficiency syndromea new term for old age?

    J Endocrinol

    (1999)
  • C.W Wilkinson

    Decreased hypothalamic–pituitary–adrenal axis sensitivity to cortisol feedback inhibition in human aging

    Neuroendocrinology

    (1997)
  • T.E Seeman et al.

    Price of adaptation—allostatic load and its health consequences

    Arch Intern Med

    (1997)
  • D.N Brindley et al.

    Possible connections between stress, diabetes, obesity, hypertension, and altered lipoprotein metabolism that may result in atherosclerosis

    Clin. Sci (Colch)

    (1989)
  • P Bjorntorp

    “Portal” adipose tissue as a generator of risk factors for cardiovascular disease and diabetes

    Atherosclerosis

    (1990)
  • R.M Sapolsky et al.

    The neuroendocrinology of stress and agingthe glucocorticoid cascade hypothesis

    Endocr Rev

    (1986)
  • S Lupien et al.

    Basal cortisol levels and cognitive defects in human aging

    J Neurosci

    (1994)
  • S.J Lupien et al.

    Cortisol levels during human aging predict hippocampal atrophy and memory deficits

    Nat Neurosci

    (1998)
  • D Michelson et al.

    Bone mineral density in women with depression

    N Engl J Med

    (1996)
  • Cited by (387)

    View all citing articles on Scopus
    View full text