Having outlined evidence for the model, we now consider its weaknesses. One limitation is that although the model suggests that childhood stress calibrates monocytes/macrophages during a sensitive period, it fails to specify when this window opens and closes. The extant research does not provide much guidance in this regard. The SES literature typically indexes exposure with indicators like parental education or occupation. Because these markers of status generally remain stable across childhood, they cannot be used to differentiate the impact of early vs. later exposures. The experience of maltreatment also tends to be fairly stable, precluding researchers from identifying sensitive periods. That said, two recent studies have begun to shed light on issues of timing. In both cases, the studies indexed SES by having respondents report on whether their parents owned vs. rented their homes each year, and found evidence to suggest that years 2–3 of life were a sensitive period for stress-related calibration of immune responses (Cohen et al., 2004
; Miller & Chen, 2007
). These findings provide some initial evidence that the toddler years are a sensitive period for stress-related influences on the immune response. Substantiating these findings should be an important priority for future research, as should examining how well they generalize to other processes depicted in the model.
Another weakness of the model is that it depicts unidirectional relations among its components. The ANS and HPA axis engage in much crosstalk, and both readily exchange signals with inflammatory cells (Sternberg, 2006
). Pro-inflammatory cytokines released peripherally can act on the CNS, and thereby evoke significant changes in behavior, as well as alterations in cognitive and affective functioning (Dantzer et al., 2008
). Revisions of the model will need to feature this crosstalk, and specify its role in linking early stress and later disease. Relatedly, there may be value in considering a multi-system approach here. Some authors have argued that stress contributes to disease by creating “imbalances” amongst systems, rather than affecting any singular biological process (McEwen & Stellar, 1993
; Thayer & Sternberg, 2006
). Our model takes the opposite perspective, treating inflammation as a mechanistic superhighway. We see value in this approach because it calls attention to a pathogenic mechanism that is both affected by early stress and common across multiple diseases of aging. However, other biological mediators almost certainly come into play during the evolution of disease, and it will be important for later versions of the model to specify the degree to which they are necessary and/or sufficient.
Notably, the model does not acknowledge a host of physical exposures that could mediate the effects of early-life stress. Being raised in a low-SES family boosts the chances a child will be exposed to pollutants, toxicants, and infections with long-term health consequences (Evans, 2004
; Wright and Subramanian, 2007
), and at least in America, receive suboptimal access to medical care. Maltreated children are likely to have these exposures as well, because their parents are unlikely to be proactive about health promotion. Thus, to fully account for the effects of early stress, future work must consider social and physical “pollutants.” Work like this has begun and shows that these exposures can have synergistic influences on disease risks (Chen et al., 2008
; Clougherty et al., 2007
; Shankardass et al., 2009
). More work on how social and physical exposures intersect, and interact, in shaping health outcomes should be an important priority in future research.
The model also fails to consider hereditary influences. There is allelic variation in many of the genes that orchestrate inflammation, regulate autonomic and endocrine discharge, and affect corticolimbic and corticostriatal signaling (Cole et al., 2010
; DeRijk and de, 2005
; McCaffery et al., 2006
). There is also growing evidence that such variation explains variability in how early stress affects mental health (Belsky and Pluess, 2009
; Binder et al., 2008
; Bradley et al., 2008
; Caspi et al., 2002
; Gillespie et al., 2009
). As this work extends to processes more relevant to physical health (Poole et al., 2006
; Wang et al., 2009
), the influences of genetic variation will need to be added to the model. But at present such moderating influences are outside the scope of our framework.
Some readers may wonder why the model does not assign a more prominent role to psychopathology. Of particular relevance in this regard is depression. Early stressors like maltreatment increase vulnerability to major depression, as well as sub-syndromal affective difficulties (Heim and Nemeroff, 2001
). Moreover, depression is sometimes accompanied by processes featured in our model, like chronic inflammation, excessive HPA discharge, and social difficulties (Coyne, 1976
; Glassman and Miller, 2007
; Howren et al., 2009
; Stetler and Miller, in press
). Despite this overlap, we do not view depression as a necessary step on the causal pathway from early stress to later disease. There is much evidence that even without lowering mood, early stress can trigger the model’s behavioral and biological features. For example, one study found that familial harshness presaged trajectories of inflammatory responding (Miller and Chen, 2010
), and did so in a manner that was completely independent of depressive symptoms. Similarly, work from the Dunedin birth-cohort found that at age 32 clinical depression was associated with greater inflammation. However, further analysis revealed that this association disappeared when childhood maltreatment was entered into equations. The authors then stratified the sample according to maltreatment (presence vs. absence) and depression (presence/absence). Participants who were maltreated and depressed had higher inflammation, relative to non-exposed controls. There was also heightened inflammation among participants exposed to maltreatment alone. However, levels of inflammation in participants who were depressed but had not been maltreated were statistically indistinguishable from controls (Danese et al., 2008
). These patterns were echoed in the 40-year study of medical students from Johns Hopkins (Kittleson et al., 2006
), which found that low childhood SES presaged CHD risk at age 50, net of depressive symptoms. In other studies like ACE, the effects of childhood adversity were attenuated, but not eliminated, by controls for depressed mood (Dong et al., 2004
). Together, these findings suggest that childhood stress brings about inflammation, and perhaps disease, through mechanisms largely independent of depression.
Of course, other mood states and/or psychiatric conditions could be important mediators here. Consistent with this view, several papers have suggested a role for broader clusters of negative emotions (Lehman et al., 2005
; Lehman et al., 2009
). However, structural equation models have revealed that such clusters provide little incremental explanatory power beyond a model that does not include them (Lehman et al., 2009
). In other papers, negative emotions have formed part of a broader “Psychosocial Functioning” construct that includes social contacts, so it is difficult to ascertain their specific contribution (Lehman et al., 2005
; Taylor et al., 2006b
). In the broader literature, there is fairly limited evidence for mood as a mediator of the health effects of SES (Matthews and Gallo, 2011
) All of that said, a fairly narrow range of affective mediators has been considered to date, and the field would benefit from a more thorough look at other candidate states (e.g., shame, anxiety) and disorders (e.g., post-traumatic stress, antisocial personality).
Finally, the model does not address the issue of resilience, or why some individuals exposed to early stress remain healthy. Data show that even among children with lengthy and severe maltreatment, only a fraction go on to develop chronic disease. In the ACE cohort, for example, only 20% of those in the most profound adversity category went on to develop CHD as adults (Dong et al., 2004
). Of course, this value could be low because not all respondents were in the age range in which CHD manifests, and the condition has a fairly low base rate. But even under conditions where all participants are exposed to a known disease-causing agent, the same pattern of resilience emerges. For instance, in the study where adults were exposed to viruses that cause the common cold, participants from low-SES backgrounds were more likely to become sick (Cohen et al., 2004
). But even among those from the lowest SES category, fewer than 50% actually manifested diagnosable symptoms of disease, suggesting that resilience was the normative outcome.
Thus, a crucial task for the next wave of research in this area will be to specify why some individuals succumb but others are protected from the health consequences of early stress. There are hints from several recent articles that maternal nurturance may be an especially influential source of resilience, capable of offsetting some of the risky hormonal, metabolic, inflammatory, and cardiovascular profiles that tend to develop in persons exposed to childhood adversity (Chen et al., in press
; Evans et al., 2007
; Fisher et al., 2000
; Luecken and Appelhans, 2006
; Miller et al., 2011
). Certain genetic variants also seem to confer protection against stress as well (Belsky and Pluess, 2009
; Binder et al., 2008
; Bradley et al., 2008
; Caspi et al., 2002
; Gillespie et al., 2009
). As work like this matures our model will need to be refined, so that it offers pathways to both resilience and vulnerability (Chen and Miller, 2011
; Cicchetti and Blender, 2006
Apart from the limitations of the model itself, it is important to recognize weaknesses in our assessment of it. Despite the fact that children are exposed to many different kinds of chronic stressors, our evaluation of the model was based only on studies of disadvantage and maltreatment. We made the assumption that these experiences shared enough common features that they could be aggregated under the rubric of chronic stress. There were many instances where this proved to be a tenable assumption – people who were exposed to either stressor early in life tended to become adults who were vigilant, mistrusting, and social isolated. They also tended to have poor health practices, high levels of inflammation, and be at risk for CHD. But there were cases where the consequences of these experiences diverged, particularly in studies of HPA axis activity. It will be important in future research to more thoroughly evaluate the similarities and differences between these stressors, and simultaneously identify the biobehavioral residue of other adversities that children face more routinely but have not been the subject of much research (e.g., chronic parental conflict, severe illness in the family). Once more data like these become available, they can be used to evaluate the model more rigorously, and guide any necessary revisions to it.