Allostatic Load and the Assessment of Cumulative Biological Risk in Biobehavioral Medicine: Challenges and Opportunities

Abstract

Allostatic load provides a useful framework for conceptualizing the multi-system physiological impact of sustained stress and its effects on health and well-being. Research across two decades shows that allostatic load indices predict health outcomes including all cause mortality, and vary with stress and related psychosocial constructs. The study by Slopen and colleagues in this issue provides an example both of the utility of the allostatic load framework, and of limitations in related literature, such as inconsistencies in conceptualization and measurement across studies, and the frequent application of cross-sectional designs. The current article describes these limitations and provides suggestions for further research to enhance the value and utility of the allostatic load framework in biobehavioral medicine research.

Allostatic load provides an integrative framework for understanding the physiological processes through which chronic stress and other sustained psychosocial factors affect health and well-being (1). This model was introduced by McEwen and Stellar (2) to describe the biologic toll exacted by prolonged activation of primary markers in the autonomic nervous system (ANS) and hypothalamic-pituitary-adrenocortical (HPA) system, as an organism attempts to maintain “allostasis” (i.e., physiological homeostasis) in the face of environmental, psychological, and behavioral challenges. The cumulative stress responses can have damaging effects on multiple downstream secondary physiological functions, thereby increasing morbidity and mortality risks, conceptualized as tertiary outcomes in the allostatic load model (1). The model recognizes that there is wide variation in physiological and health consequences of chronic stress as a function of interacting genetic, environmental, and individual influences (3,4). In contrast to the common practice of examining risk factors within a single physiological system, the allostatic load framework provides an integrative approach that may better characterize the cumulative impact of dynamic and nonlinear influences across major biological regulatory systems.

Several recent literature reviews summarizing nearly two decades of research have concluded that allostatic load predicts health outcomes including cardiovascular disease, functional decline, frailty, and all-cause mortality (5-7). The model has also proven useful in elucidating the physiological consequences of psychosocial and socioeconomic antecedents of stress and their implications for health disparities (5,6,8,9). Although early allostatic load studies were conducted in a single cohort with limited socio-demographic variability (10,11), subsequent research has examined diverse populations and varied social constructs (e.g., socioeconomic status, immigration) (5,6). This work has strengthened the evidence for the allostatic load framework and its utility in understanding health and social correlates therein (5,6).

In the current issue, Slopen and colleagues report associations between childhood adversity and allostatic load—here, termed “cumulative biological risk”—in 550 participants from the Chicago Community Adult Health Study (12). They found that participants who reported experiencing greater adversity in childhood had increased dysregulation across physiological systems, but only if they also resided as adults in neighborhoods characterized by low affluence (operationalized using census data). The authors concluded that the resources inherent to an affluent environment could buffer the harmful physiological consequences of early life adversity. Through this application of the allostatic load framework, the study provides a unique contribution towards understanding the lifecourse impact of early stress exposure on a range of deleterious physiological outcomes, as moderated by neighborhood context.

The study also highlights several limitations of the extant allostatic load literature that deserve further consideration. In particular, the research provides an example of unsettled questions regarding the optimal representation of allostatic load (5,6). Allostatic load is typically operationalized as a composite of biological markers representing multiple systems, especially the neuroendocrine, cardiovascular, metabolic, and immune systems. Allostatic load composite scores often combine primary mediators of the stress response (e.g., stress hormones; pro-inflammatory cytokines) and secondary outcomes of cardiovascular, metabolic, and immune dysregulation (e.g., blood pressure, waist circumference, glycosylated hemoglobin), measured at a single point in time. However, the research base is notable for the substantial variability in the specific indicators chosen, the number of indicators used both across and within physiological systems, whether continuous or categorical scores are combined, how categorical thresholds are derived, whether or not the biomarkers are adjusted for use of medications, and how indicators are combined into composite scores (e.g., whether or not a weighted approach is used) [for a detailed description of approaches used to operationalize allostatic load, see (5)]. A common method is to form a simple composite by counting the number of physiological parameters at relatively “high” or “low” levels based on the distribution of scores in the study sample (e.g., extreme quartile approach), either overall or specific to sex groups. However, because the distributions depend on the population, this method compromises uniformity across studies. Moreover, sample-based thresholds may be useful for research purposes, but their psychosocial, medical, and public health applications are less clear, and the statistical underpinnings of these composite scores generally are not well documented. To bypass this concern, some studies have instead applied a priori clinically defined thresholds to categorize participants as high or low risk. However, not all physiological components selected to define allostatic load have known clinical thresholds, including the primary neuroendocrine and inflammatory biomarkers. Other investigators have tested complex scoring methods to create allostatic load indices, for example, relying on recursive partitioning (13), canonical correlation (14), and “grade of membership” (15,16) multivariate models. The limited comparative research suggests that the choice of indicators and the approach used to combine them has only a modest bearing on predictive utility (4,13,14), although maintaining the continuum of scores and/or incorporating both high and low extremes may enhance the ability to detect associations of allostatic load with a range of health outcomes (5,15,17,18).

The study by Slopen and colleagues applied a composite of eight cardiovascular, metabolic, and immune system indicators with known clinically relevant thresholds. Notably, none of the primary biomarkers were included in the composite, and this may be a reason that the authors applied the term “cumulative biological risk”, instead of allostatic load. This practice is relatively common, even though the primary mediators are considered a fundamental element in the allostatic load framework. The approach to conceptualizing allostatic load in large epidemiological studies is often determined pragmatically according to the available information, and most of these trials (for example, the National Health and Nutrition Surveys) do not include neuroendrocrine assessments. Moreover, the primary biomarkers are difficult to measure consistently, as they may be collected from varied biological samples (e.g., blood, saliva, urine) and are prone to circadian fluctuations as well as rapid responses from external stimuli (8,19). Thus, a single point-in-time assessment may not adequately capture true dysregulation. Also notable is the fact that several of the indicators used in the study by Slopen overlap with those used to diagnose the metabolic syndrome (i.e., waist circumference, systolic and diastolic blood pressure, high-density lipoprotein cholesterol, glycosylated hemoglobin, which captures similar information to fasting glucose) (20). The similarity between allostatic load and the metabolic syndrome, or other composite systems such as the Framingham Risk Score, has been noted previously (11). There is some evidence that metabolic syndrome and allostatic load are conceptually distinct (21) and that the variance in health predicted by allostatic load is greater than that of the metabolic syndrome (14,22). Nonetheless, additional research to better understand the relative predictive utility of these and related composite frameworks is indicated. In addition, it remains important for investigators to describe findings for individual components of allostatic load, as Slopen and colleagues have done, to enhance comparability across studies and identify the specific components of allostatic load that contribute to results based on composite scores.

Like much prior research, the study by Slopen et al. also applied a cross-sectional design, creating unanswered questions about the directionality and temporal progression of observed associations. The allostatic load model fundamentally proposes that the physiological burden of stress is cumulative over time—a premise that cannot explicitly be examined within a cross-sectional design. Even longitudinal studies may inadequately address this central tenet if the follow-up period is limited, since duration of exposure necessary for primary mediators to induce secondary and tertiary outcomes is unknown. Similarly, a cross-sectional design prevents determination of whether or not the secondary physiological dysregulation is in fact subsequent to primary mediators and to the stress exposure. Reverse-causation or loop mechanisms may also be relevant, because individuals with greater physiological dysregulation could experience, and thus report, higher or added stressors at the point of mutual assessment (22). These complex associations cannot directly be explored through a cross-sectional study with a single assessment occasion. The current Slopen study circumvents this concern to some extent by (retrospectively) assessing previous childhood experience as exposure. Nonetheless, only a few prospective studies to date support the applicability of the allostatic load framework (23,24) and the need for additional longitudinal research has been repeatedly emphasized (5,25,26).

Although research suggests that social and health correlates are generally consistent across different operationalizations of allostatic load, definitional variability may impede the adoption and application of the allostatic load framework in research and practice. The absence of a “gold standard” representation also limits the comparisons that can be made across applications. Such variability is common in the early stages of biomarker panel development (19), but to further advance the field, agreement is needed regarding the biological indicators and systems that should optimally be used to operationalize allostatic load (5,19). Moreover, additional observational studies in humans and targeted experimental investigations are warranted to evaluate the health implications of allostatic load, especially in long-term prospective cohorts with data on interacting factors (genetic influences, health behaviors). Such studies will improve understanding of the lifecourse dynamics of allostatic load-related processes in health and disease (25-27). Further research is also needed in diverse samples to investigate the consistency of observed findings, and the utility of specific allostatic load operationalizations, across different populations. In sum, the allostatic load framework and its potential for advancing biobehavioral medicine research and practice will be enhanced by efforts to standardize operationalization, disentangle the temporal sequence among hypothesized antecedents, mediators, and physiological consequences, and to examine the generalizability of these and other advances in allostatic load across diverse groups.