Stress system reactivity moderates the association between cumulative risk and children’s externalizing symptoms

Abstract

This study examined children’s stress system reactivity via the autonomic nervous system (ANS) and hypothalamic pituitary adrenal axis (HPAA) during an acute stressor as moderators of predicted relations between cumulative risk (CR) and mental health symptoms in a sociodemographically diverse sample of young children (n = 58). We employed a reliable stressor paradigm to allow assessment of individual differences in respiratory sinus arrhythmia (RSA) and pre-ejection period (PEP), indexing ANS reactivity, and salivary cortisol, indexing HPAA reactivity. Children’s behaviours were assessed using the Child Behaviour Checklist (CBCL). Cumulative risk was indexed by eight parent-reported sociodemographic and psychosocial risk factors. There was a significant main effect of CR on externalizing but not internalizing behaviours. Significant moderations emerged showing that among children with high CR, less RSA withdrawal during the acute stressor and less cortisol recovery following the stressor were associated with to greater externalizing behaviours. Among children with low CR, RSA and cortisol recovery were unrelated to internalizing or externalizing symptoms. Cortisol and PEP reactivity were not significant moderators. Results highlight the relevance of stress system function for understanding differential susceptibility to the early emergence of externalizing symptoms, linked to cumulative risk exposure.

1. Introduction

Repeated exposure to stress increases children’s risk for emotional and behavioural dysregulation, particularly when the effects of cumulative adversity are considered (Shonkoff et al., 2012; Luthar, 2006). The Adverse Childhood Experience Study provided seminal evidence in understanding the importance of childhood stressors, including exposure to abuse, neglect, violence, family turmoil, and family separation, in elevating children’s risk for adverse mental health outcomes (Dube et al., 2001). More recent work has examined these experiences alongside poverty-related stressors in a cumulative risk (CR) model to predict the emergence of mental health symptoms (Atkinson et al., 2015; Evans et al., 2013; Raviv et al., 2010; Atzaba-Poria et al., 2004). Findings across these studies are consistent with early CR models suggesting that the accumulation of risk factors early in life is more predictive of negative outcomes than any single risk factor (Rutter, 1979). In a 20-year longitudinal study, childhood CR was linked to broad range of adverse outcomes across development, showing that 1) as the number of risk factors increased, the number of negative outcomes increased and 2) CR predicted negative effects on cognitive, quality of life, and mental health outcomes from preschool age to adulthood (Atkinson et al., 2015). Despite the demonstrated impacts of CR, there is a more limited understanding of why different children exhibit variable manifestations of mental health and behavioural challenges after similar stressful experiences. Children who experience high cumulative risk are not equally susceptible to adverse outcomes, and individual differences in stress-system vulnerability have been proposed as a relevant factor for understanding the likelihood of healthy outcomes.

Early childhood represents a critical developmental period when neural plasticity is at its peak and environmental factors are highly influential on the long-term adaptive functioning of stress physiology (Davidson, 2000). Early adverse environments can disrupt normal development of stress systems and consequently constrain regulatory capacities for coping with and adapting to fluctuating contexts. Two distinct theoretical frameworks have been proposed as possible explanations for individual differences in vulnerability to risk and later maladaptation. According to differential susceptibility theory, varying behavioural, physiological, or genetic characteristics differentially impact children’s vulnerability to maladaptation from environmental adversity (Ellis et al., 2011); some children are more likely to experience adverse effects from stress while others are resilient (Masten and Obradović, 2006). The biological sensitivity to context framework suggests that children with heightened sensitivity to environmental demands are more susceptible to the adverse effects of cumulative adversities but may actually benefit the most from positive contexts (Boyce and Ellis, 2005). While both models predict that some children are more susceptible to adversity than others, there is still a limited understanding of the particular biological mechanisms that contribute to individual differences in vulnerability to adversity and associated psychological dysfunction in early childhood. Identifying the early environmental factors that confer risk for maladaptive functioning and dysregulated stress physiology will provide insight into a possible window of opportunity during which stress system function is still malleable and amenable to positive intervention.

Here we address stress system function as a potential vulnerability factor for early emerging mental health symptoms related to cumulative risk by examining children’s stress reactivity during an acute stressor. Examining this question in young children may hold particular value because (1) the early emergence of problem behaviours is a strong predictor of later life challenges (Campbell et al., 2000) including lower social competence (Chang et al., 2012), worse academic achievement (Swanson et al., 2012), and more mental and physical health problems (Atkinson et al., 2015) and (2) early identification of most at-risk children offers a unique opportunity for impactful intervention given rapid self-regulatory development and environmental sensitivity in the preschool years.

1.1. Cumulative risk and mental health symptoms in early childhood

The cumulative risk theory posits that as the number of risk factors increase, the greater the prevalence of clinical problems (Appleyard et al., 2005). Factors such as maltreatment, family turmoil, poverty, and family disruption are included in CR and have been shown to have a negative impact on child development. For example, exposure to maltreatment, domestic violence, and family stress at varying timepoints in childhood predicted higher levels of anxiety and depression during adolescence (Appleyard et al., 2005). The preschool years are a sensitive period for the development of emotional and behavioural self-regulation, so early negative experiences may be particularly significant in conferring risk for externalizing problems. Because mental health problems emerging in early childhood strongly predict future psychopathology (Atkinson et al., 2015), understanding how risk leads to maladaptation may be valuable in informing early intervention for promoting resilience. Multiple studies have consistently demonstrated associations between CR and internalizing and externalizing symptoms, with associated problem behaviours emerging as early as three years (Gutman et al., 2019). The vast majority of work examining the impact of CR have corroborated early work that the additive effect of risk factors more strongly predicts maladaptive outcomes than individual risk factors (Atkinson et al., 2015; Raviv et al., 2010; Slopen et al., 2014; Solomon et al., 2016; Trentacosta et al., 2008). While cumulative risk reliably predicts maladaptive outcomes, children are not simply products of their environment, and research has implicated multiple biological vulnerability factors that may explain why some children are more susceptible than others to psychopathology related to early adversity.

1.2. Stress neurobiology

Individual differences in stress system reactivity may provide important information on children’s susceptibility to the effects of cumulative risk on mental health and behaviour. The functional role of these systems is to enable the rapid engagement and disengagement of arousal, attention, and motor control to effectively respond to changing environmental demands and return to homeostasis after immediate demands have passed. Accordingly, their activation may reflect a variety of inter-related processes such as appraising the needs of a given situation, then effectively mobilizing relevant resources and behavioural approaches. Children may exhibit emotional and behavioural vulnerability culminating in mental health difficulties, should they have challenges in either appraising needs or carrying out effecting responses. For example, a child with relatively high CR exposure may have more impulsive arousal to control, which may interfere with their ability to effectively appraise and respond to a situation appropriately resulting in poor behavioural and emotional self-regulation (Roos et al., 2017). Here we try to elucidate the intervening physiological mechanisms involved in children’s ability to respond to situations requiring effortful control of arousal and attention, and under which circumstances these abilities become dysregulated to predict behavioural and emotional difficulties.

Research on children’s responses to stress has highlighted the importance of considering both the autonomic nervous system (ANS) and the hypothalamic-pituitary-adrenal axis (HPAA) to more fully capture the underlying mechanisms involved in regulating responses to contextual environmental demands. Early childhood represents a sensitive period for rapid development of the ANS and HPAA (Rutter, 1991) and may be key to understanding when early stressful experiences exert maximal effects and confer risk for later psychopathology.

1.2.1. Autonomic nervous system

The autonomic nervous system (ANS) may be an important underlying mechanism in the link between CR and early emerging problem behaviours, given its role in regulating emotional, behavioural, and physiological responses to stress (Obradović, 2016). Generally, an adaptive autonomic response to acute stress is characterized by flexible engagement and disengagement of arousal and rapid recovery following the stressor, which facilitates appropriate attentional, emotional, and behavioural response to guide effective coping (Porges et al., 1994). Relative to the literature on autonomic reactivity, very few studies have examined autonomic recovery; however, there is a general assertion that under adverse family environments, greater physiological arousal and slower rates of recovery are associated with more behavioural and emotional problems in young children (Miller et al., 2013; Rudd et al., 2017).

The parasympathetic nervous system (PNS) is one branch of the ANS with the primary role of regulating heart rate via the vagus nerve, allowing flexible physiological responses and indirectly modulating behavioural demands (Yim et al., 2015; Thayer et al., 2009; Thayer et al., 2012). While PNS activation promotes rest in the absence of immediate threat, PNS suppression/withdrawal in the context of challenges is critically important for increasing attention and effectively mobilizing adaptive resources for coping with environmental demands (Porges, 2001). Respiratory sinus arrhythmia (RSA) serves as an index of PNS function, specifically reflecting parasympathetic regulation of the heart rate. RSA has been shown to increase with age and become stable through early childhood and potentially up to the age of seven (Alkon et al., 2006; Calkins and Keane, 2004); however, RSA trajectories are less consistent across studies with older children (Alkon et al., 2003; El-Sheikh, 2005). Variability in measures of RSA during acute stress has shown to be predictive of differences in young children’s mental health outcomes (Calkins et al., 2007; Hinnant and El-Sheikh, 2009).

There is consistency across findings that low reactivity to challenges (indexed by less RSA withdrawal) is associated with problem behaviours and poor emotion-regulation in children whereas high reactivity (indexed by greater RSA withdrawal) is related to better attention and emotion-regulation (Calkins et al., 2007; Obradović et al., 2010; Boyce et al., 2001). However, since RSA reactivity is largely dependent on task demands, more emotionally evocative tasks have been shown to elicit greater RSA withdrawal that predicts a mixed profile of internalizing and externalizing symptoms (Calkins et al., 2007). When examining symptoms independently, children with predominantly externalizing symptoms demonstrated less RSA withdrawal. Researchers have proposed several reasons for these seemingly contradictory findings suggesting that normative and clinical samples show varying responses such that low RSA reactivity to stressors is related to externalizing problems in normative samples (Boyce et al., 2001; Calkins et al., 2007) but high RSA reactivity is more consistent with clinically relevant emotional and behaviour problems (Crowell et al., 2005). Nonetheless, these findings across studies suggest that, in general, young children that withdraw parasympathetic control of heart rate during challenges may be able to more flexibly engage adaptive behavioural and emotional responses. Another important mechanism to consider is the role of the sympathetic nervous system (SNS), the second branch of the ANS which is engaged in response to a perceived threat and prepares the body for a fight-or-flight response.

SNS activation results in a more prolonged state of arousal that is more difficult and metabolically costly to regulate in comparison to the PNS (Beauchaine, 2001; Porges et al., 1994). SNS activity is typically indexed via pre-ejection period (PEP), the time between when the heart fills with blood and when blood is ejected (with shorter PEP indicating SNS activation; Porges et al., 1994). Measuring PEP under laboratory conditions can often be challenging with young children because it is highly sensitive to motion artifacts; accordingly, the SNS has garnered much less attention in the literature, despite evidence that SNS reactivity is linked to internalizing and externalizing symptoms in children facing family adversity (e.g., El-Sheikh, 2005). Elucidating the role of SNS in the context of cumulative risk exposure may be particularly relevant for understanding individual differences in ANS reactivity more generally, but also for understanding differences in PNS regulation. The two autonomic branches are not mutually exclusive and show similar developmental trajectories; as with measures of PNS, PEP indexing SNS activity increases in childhood and tends to show moderate stability into adolescence (Matthews, 2002). Considering that interactions between the two autonomic branches have been implicated in children’s risk for maladaptive behavioural outcomes in the context of a stressful home environment (El-Sheikh et al., 2009), there is a need to examine patterns of physiological regulation across multiple systems in order to more holistically and meaningfully identify biomarkers of vulnerability for maladaptive outcomes under varying degrees of risk exposure (Bauer et al., 2002; Giuliano et al., 2018; Gatzke-Kopp and Ram, 2018).

1.2.2. Hypothalamic-pituitary adrenal axis (HPAA)

The HPAA modulates levels of the glucocorticoid cortisol in the blood stream to maintain homeostasis in the face of perceived threat or stress. Although the HPA axis shows a protracted development and does not reach full maturity until adolescence, excessive stress exposure in early childhood can cause dysregulation of HPA function resulting in heightened or prolonged activation during stressful situations (Gunnar and Donzella, 2002). Several studies have implicated HPAA dysregulation as a vulnerability factor for children’s emotional and behaviour problem with mixed findings. Both blunted (Sturge-Apple et al., 2012) and elevated (Barrios et al., 2017) cortisol responses have been observed in young children who have experienced early adversity. Similarly, both high (Hatzinger et al., 2007) and low (Von Klitzing et al., 2012) HPAA reactivity have been linked to greater internalizing and externalizing behaviours and that HPAA dysregulation likely functions to moderate the impact of early adversity on the emergence of mental health symptoms (McEwen, 2007). Surprisingly few studies have examined the role of cortisol recovery in the development of mental health symptoms, particularly with a young sample as in the present study. Some prior work has linked blunted cortisol recovery to greater internalizing and externalizing symptoms, but this was in older children and adolescents (Nederhof et al., 2015; Schoorl et al., 2017). Another recent study found that slow HPA recovery predicted greater adolescent externalizing symptoms, but the interaction of other physiological systems qualified this finding (Wadsworth et al., 2019). It is presumed that delayed or blunted HPAA recovery reflects an inability to effectively regulate physiology and return to baseline levels following stress and this may be particularly important for understanding children’s ability to adaptively employ various coping behaviours in the face of challenges. The present study aims to distinguish the roles of cortisol reactivity and recovery in predicting children’s internalizing and externalizing behaviours related to early adversity.

1.2.3. Moderating role of stress neurobiology

It is well established that early environmental adversity has deleterious effects on children’s development. Similarly, a growing body of research has implicated dysregulation of stress neurobiology in the development of psychopathology. However, much less is known about how adversity and stress system functioning interact to differentially impact children’s behavioural and emotional outcomes. It has been proposed that physiological reactivity is an intervening biological mechanism that moderates the effect of early stressors on behavioural and emotional outcomes (McEwen, 2007). Recent studies have found that children who have been exposed to repeated family stress and have dysregulated ANS (Obradović et al., 2010; El-Sheikh et al., 2008; Erath et al., 2009) and HPAA functioning (Badanes et al., 2011; Barrios et al., 2017) are more vulnerable for developing internalizing and externalizing problems. Research has started to focus on vulnerability stress models, suggesting that stress system dysregulation is a vulnerability factor which leaves dysregulated individuals more susceptible to maladaptive outcomes following exposure to adverse environments (Belsky, 2005; Boyce and Ellis, 2005) including harsh parenting and peer victimization (von Klitzing et al., 2012). However, still very few studies have examined RSA as a vulnerability or protective factor in children with varying degrees of adversity or within a cumulative risk framework. Some studies have shown that children with high RSA reactivity (greater RSA withdrawal) to emotionally evocative stimuli were buffered from the adverse effects of marital conflict on problem behaviours and that children with low RSA reactivity (less RSA withdrawal) were more at risk for maladaptive outcomes (El-Sheikh et al., 2001). Similarly, some studies have implicated HPAA dysregulation as a vulnerability marker such that high cortisol reactivity in the context of high family stress and harsh parenting predicted more emotional (von Klitzing et al., 2012) and behaviour problems (Obradović et al., 2010; Barrios et al., 2017). However, there is still a paucity of studies examining cortisol reactivity during age-appropriate laboratory stressors, and even fewer that consider cortisol recovery as a potential vulnerability factor.

There are still key questions regarding the influence of stress system functioning on children’s emotional and behavioural problems, particularly in those experiencing varying degrees of cumulative stress. The majority of studies examining the interaction of early adversity and stress system function on adaptive and maladaptive outcomes have only considered individual risk factors (e.g., harsh parenting, marital conflict); however, given that cumulative risk is consistently shown to be a more reliable predictor of adverse physiological and behavioural outcomes, it seems necessary to evaluate this moderation under varying levels of cumulative risk. There are obvious gaps in the field that need to be addressed to better understand how early adversity interacts with physiology to either buffer or protect children from the adverse outcomes associated with stress exposure. We extend the literature by examining both ANS and HPAA function as potential moderators in the link between CR and young children’s mental health symptoms.

1.3. The present study

In the present study, we examined how stress system reactivity interacts with CR to predict early mental health symptoms in a sample of four to six year-old children. Stress system reactivity was examined via the ANS and HPAA during a validated laboratory acute matching stressor task (Kryski et al., 2011; Tolep and Dougherty, 2014). We expected that both ANS and HPAA reactivity would moderate the link between CR and internalizing and externalizing symptoms. This work will provide greater specificity in understanding why some children exposed to multiple stressors may be particularly vulnerable to psychopathology as well as inform interventions for adversity-exposed children and their caregivers to enhance stress-management abilities as a way to promote resilient trajectories for mental and behavioural health.

2. Methods

2.1. Participants

Mother-child dyads volunteered to participate through community recruitment in a mid-sized city in the United States as part of a larger study (Roos et al., 2017; Roos et al., 2019). Dyads were eligible to participate if children had no history of psychiatric disorders, developmental delays or serious health problems. Participants were randomly assigned to a laboratory control or stressor condition (only the stressor condition presented in this study). Participants were 58 preschool children ranging in age from 4 to 6 years (M = 5.32, SD = 0.65). The majority of children were female (56.9%), White (65.5%) and lived in two-parent households (48.3%). The median maternal education was “some college or associate’s degree” and the median reported annual household was $30,000–$39,999 (range <$4999–$100,000+).

2.2. Procedure

Participants completed all measures in a single 2-hour laboratory visit scheduled to start between 9 am and 3 pm. Children awoke at least 1 h prior and did not eat at least 1 h prior to the visit. Upon arrival, an assessor engaged with the child and began a developmentally appropriate explanation of the activities using a sticker chart and colouring book. After the description of laboratory activities, children gave verbal assent and mothers gave written consent to participate. Next, mothers assisted the assessors in ensuring child comfort with application of electrodes for monitoring ANS physiology.

To assess baseline ANS physiology, mother and child watched a 5-minute relaxing ocean video while sitting across a table from each other but not touching (Piferi et al., 2000). Saliva samples were collected at five points throughout the visit, with sample 2 reflecting an in- lab baseline (as reported in Roos et al., 2017). Mothers completed questionnaires in a separate room while the child completed various tasks including a stressful matching task, which is the focus of this study. ANS measures were collected throughout the matching task to examine effects of stress reactivity relative to children’s baseline ANS measures.

2.3. Measures

2.3.1. Acute stressor manipulation

The stressor employed in this study was a matching task adapted from previous work (Kryski et al., 2011) and has been established to invoke cortisol, RSA, and PEP reactivity in young children compared to a control condition (Roos et al., 2017). Children played a matching game in which they were given a colour-coded legend and were instructed to match coloured stickers to different transportation types displayed on a worksheet with 30 images contained in squares. Before beginning, the child chose a prize to earn for successfully finishing the worksheet. Children were required to complete the worksheet in two-minutes, an insufficient amount of time for most 4- to 6-year-old children. Children had three attempts (trials) to win their prize. An unfamiliar assessor with a stern, flat affect operated a stop-light timer which showed a green-light for 90 s, followed by a yellow light for 30 s, then a red light accompanied by a loud beep signalling the end of the allotted time. During each trial, the assessor gave negative feedback at least twice (e.g., “You’ll need to go faster.”). After three failed trials, the assessor told the child they did not earn their prize, left the room, and a friendly, familiar assessor returned. After all tasks were complete, the child was debriefed about the nature of the matching task and told they did an excellent job and had earned their desired prize.

2.3.2. Cortisol

Salivettes were used to collect saliva samples (Sarstedt, Inc., Newton, NC) that were frozen (−20 °C) and sent to University of Trier for assaying. Five total samples were collected (baseline at lab entry and 0, 20, 40, and 50 min post matching task). Six participants were excluded from analysis due to refusal or compromised saliva samples (e.g., participant ate during visit, slammed finger in door, did not produce enough saliva). Extreme values (>3SDs condition mean) were winsorized.

2.3.3. Autonomic physiology

To assess ANS physiology, eleven electrodes were utilized to measure pre-ejection period (PEP), indexing SNS activity, and respiratory sinus arrhythmia (RSA), indexing PNS activity. Electrocardiogram (ECG) was obtained at sampling rate of 500 Hz via three pre-gelled electrodes applied in a lead II arrangement on the distal end of the right clavicle, lower left rib cage, and the lower abdomen. Eight electrodes were applied in a tetrapolar formation on the left and right lateral neck and torso, from the jawline to the diaphragm, to measure impedance cardiography (ICG, or “Z0”). Data was collected wirelessly with Biopac transmitters (Biopac Systems Inc., Goleta, CA) sending ECG and ICG signals to a Biopac MP150 acquisition unit located in the room with the participant. Data was processed using Mindware HRV and IMP software (Gahanna, OH). ECG signals were visually inspected by trained research assistants to confirm heart beats in 30 second epochs and to confirm Q and B placement for processing PEP. RSA values were estimated from the power in the respiratory frequency band (0.24–1.04 HZ) derived from the spectral density function. RSA values were then averaged across 30 second epochs to derive a baseline RSA value and a task RSA value. The 30s epochs were averaged across baseline, first-half matching task and second-half matching task, as higher stress was predicted to occur during the second-half of the matching task as children experienced repeated failure. PEP was indexed from the derivative of the cardiovascular impedance signal (dZ/dt) and calculated as the length of time from the Q-point of the ECG waveform to the B-point of the dZ/dt waveform. Among participants with usable behavioural data, participants were excluded from analyses due to refusal of electrodes or having less than 50% artifact-free data for baseline measure or second- half of the matching task (RSA, n = 11; PEP, n = 14). Missing autonomic data was not associated with any inferential variables of interest (all ps > 0.10).

2.3.4. Child behaviour checklist

Mothers reported children’s behavioural and emotional problems using the Child Behaviour Checklist (CBCL; Achenbach and Rescorla, 2000), which consists of 100 items on a 3-point Likert scale (0 = not true, 1 = somewhat/sometimes true, 3 = very/often true). The CBCL generates scores on 9 psychopathology subscales to yield an internalizing, externalizing, and total problem score. The internalizing problem score is derived from the emotionally reactive, anxious/depressed, somatic complaints and withdrawn subscales. The externalizing problem score is derived from the attention problems and aggressive behaviour subscales. The CBCL has been widely used and has excellent psychometric properties (α = 0.92; Achenbach, 2011). For the purpose of this study, internalizing and externalizing scores were the focus of our analyses.

2.3.5. Cumulative risk profile

Stress exposure was indexed via a composite measure of cumulative risk using eight risk factors across sociodemographic and psychosocial domains. Each risk factor was coded dichotomously as 0 for not present and 1 for present. Dichotomized scores were summed (unweighted) to compute a total CR score. Three measures of sociodemographic risk included low-household income (<$30,000), single-parent status, and maternal education below high school. In addition, psychosocial risk factors were measured using four subscales of the Life Experiences Survey (LES; Sarason et al., 1978) including exposure to family turmoil, child-family separation event, exposure to violence, and poverty-related stressors. Each subscale of the LEC includes several composite questions, and risk was considered present if mean scores on each subscale were above the upper quartile. Chaos and disorganization in the child’s home environment was assessed using the Confusion, Hubbub, and Order Scale (CHAOS; Matheny et al., 1995) with scores in the upper quartile indicating risk was present. Composite scores of CR including all eight possible risk factors were calculated for 54 participants. Cumulative risk scores were not calculated for any participants with missing data from any of the eight risk factors (n = 4). Composite risk scores fell within a range of 0 to 5 (M = 1.48; SD = 1.37; Median = 1). Frequencies are presented in Table 1.

Table 1.

Descriptive data.

	n	%	Mean	(SD)
Child demographics	58		5.32	(0.63)
Age
Sex
Female	33	43.1
Male	25	56.9
Ethnicity
White	38	65.5
American/Black	3	5.2
Hispanic/Latino	7	12.1
Mixed	6	10.3
Unknown	1	1.7
Missing	3	5.2
Cumulative risk	54	93.1	1.48	(1.37)
Frequencies
0	14	24.1
1	19	32.8
2	10	17.2
3	5	8.6
4	4	6.9
5	2	3.4
Child outcomes
Cortisol Peak	52		0.68	(1.50)
Cortisol Recovery	52		−0.40	(1.26)
PEP	44		−3.17	(9.21)
RSA	47		−1.06	(0.75)
CBCL Internalizing Behaviour T Score	55		52.38	(9.81)
CBCL Externalizing	55		48.29	(9.27)

Behaviour T score.

2.3.6. Data analytic plan

All analyses were conducted in SPSS. Preliminary correlational analyses were conducted to determine if child age and sex were predictive of internalizing and externalizing behaviours to determine any covariates to control for in subsequent analyses. Correlations among all variables are reported in Table 3. Next, bivariate correlations among all variables were conducted, with particular interest in the association between composite cumulative risk scores and internalizing and externalizing problem scores. Subsequent analyses examined RSA, PEP, cortisol reactivity, and cortisol recovery separately as potential moderators between cumulative risk scores and both internalizing and externalizing problem scores using Hayes’ PROCESS add- on in SPSS v.22 (Hayes, 2013). Cortisol, PEP, and RSA reactivity were calculated by subtracting baseline from peak reactivity (i.e., highest cortisol, 20 min post-stressor; shortest PEP and lowest RSA, second-half matching task). Cortisol recovery was calculated by subtracting peak reactivity from final sample (i.e., 50 min post-stressor).¹ Descriptive statistics for biological measures across time are presented in Table 2.

Table 3.

Zero-order correlations between all variables.

	1	2	3	4	5	6	7	8	9	10
1. Child Age	–
2. Child Sex	0.182	–
3. Child Race	−0.224	−0.130	–
4. Cumulative Risk	0.047	0.028	−0.122	–
5. CBCL Internalizing T Score	0.232	0.002	0.067	0.204	–
6. CBCL Externalizing T Score	0.029	−0.059	0.072	0.321*	0.683**	–
7. Cortisol Peak	0.091	0.128	−0.072	−0.143	0.031	−0.175	–
8. Cortisol Recovery	−0.069	−0.069	0.002	0.256	−0.088	0.209	−0.676**	–
9. PEP	0.159	−0.054	−0.071	0.169	0.096	0.223	−0.451**	0.326*	–
10. RSA	−0.006	0.098	0.193	0.202	0.189	0.368*	−0.314*	0.446**	0.396**	–

^*Correlation is significant at the 0.05 level (2-tailed).

^**Correlation is significant at the 0.01 level (2-tailed).

Table 2.

Descriptive statistics for biological measures across time.

	M(SD)
Cortisol (nmol/L)
Lab entry	2.42 (1.11)
0 min post MT	1.72 (1.01
20 min post MT	2.40 (2.00)
40 min post MT	2.15 (1.53)
50 min post MT	2.00 (1.31)
RSA (power)
Baseline	6.63 (1.12)
MT first-half	5.92 (1.08)
MT second-half	5.57 (1.20)
PEP (ms)
Baseline	90.65 (10.12)
MT first-half	90.25 (10.60)
MT second-half	87.48 (10.41)

Note: nmol/L = nanomole per liter; ms = milliseconds. Roos et al., 2017.

3. Results

3.1. Correlations among variables

Zero-order correlations among all variables of interest and covariates of child age, gender, and race and were conducted (Table 3). Age, gender, and race were not significantly correlated with any variables of interest and were not included in subsequent analyses (ps > 0.10). Cumulative risk was positively correlated with externalizing symptoms [r(52) = 0.321, p = .018] such that as the number of risk factor increased, children’s externalizing symptoms increased (Fig. 1). RSA was positively correlated with externalizing behaviours [r(42) = 0.368, p = .014] such that children with higher RSA values, compared to baseline, (i.e., less RSA withdrawal/blunted reactivity to stress) had greater parent-reported externalizing symptoms (Fig. 1). Cumulative risk and RSA were not correlated with internalizing symptoms (ps > 0.10). Cortisol reactivity and cortisol recovery were not correlated with either internalizing or externalizing symptoms (ps > 0.10). All descriptive statistics are reported in Table 1.

Fig. 1.

Children’s externalizing behaviours by total number of risk factors

3.2. Moderation analyses

Hayes (2013) PROCESS macro for SPSS was utilized to examine the relationship between CR and child internalizing and externalizing symptoms as moderated by stress-system engagement (RSA reactivity, PEP reactivity, cortisol reactivity, and cortisol recovery). Regarding autonomic nervous system reactivity, there was a significant interaction between RSA and CR on children’s externalizing (β = 2.72, SE = 1.21 p = .03, R = 0.28; see Fig. 2)² but not internalizing symptoms (β = 0.15, SE = 1.47, p > .10, R² = 0.06). We examined the interaction two-ways to inform the interpretation of results, given multiple relevant theoretical perspectives.

Fig. 2.

Interactions between the number of risk factors

Note: Number of risk factors and stress system reactivity are shown and ± 1 SD from the mean for visualization of the reported moderation effects.

* significant at p < .05

** significant at p < .01.

In considering the moderating effects of CR on the link between RSA and externalizing, blunted RSA reactivity (i.e., higher RSA values) to the stressor was associated with more externalizing symptoms for children with average (M = 1.53; β = 4.25, SE = 1.65, p = .014) and higher CR (i.e., 1SD above mean = 2.92; β = 8.02, SE = 2.30, p = .001). In children with low CR (i.e., 1SD below mean = 0.15), RSA was unrelated to mental health symptoms (β = 0.47, SE = 2.40, p > .10; see Fig. 2a).

In considering the moderating effects of PNS reactivity on the link between CR and externalizing behaviours, CR was only associated with more externalizing problems for children with blunted RSA reactivity (i.e. 1SD above mean = −0.31; β = 3.14, SE = 1.30, p = .02; see Fig. 2b). However, children with average (M = − 1.09; β = 1.03, SE = 0.92, p >.10) or high RSA reactivity (i.e. 1SD below the mean = −1.87; β = −1.09, SE = 1.34, p > .10) did not exhibit elevated externalizing symptoms in the context of high CR.³ There were no associations between PEP reactivity and externalizing symptoms.⁴

Regarding cortisol recovery, there was a marginally significant interaction between CR on children’s externalizing symptoms (β = 2.11, SE = 1.05, p = .050, R² = 0.20).⁵ In considering the moderating effect of CR on the link between cortisol recovery and externalizing problems, blunted cortisol recovery following the stressor was associated with more externalizing symptoms for children with high CR (β = 5.23, SE = 2.32, p = .03; Fig. 3a). However, for children with low (i.e., 1SD below mean = 0.09; β = −0.77, SE = 1.31, p > .10) or average CR (i.e., M = 1.51; β = 2.23, SE = 1.15, p > .05, cortisol recovery was unrelated to mental health symptoms. In considering the moderating effect of cortisol recovery on the link between CR and externalizing problems, CR was associated with more externalizing problems only for children with blunted cortisol recovery (i.e., 1SD above mean = 0.90; β = 4.17, SE = 1.45, p = .006; see Fig. 3b). However, for children with average (M = − 0.40; β = 1.41, SE = 0.92; p >10) and faster cortisol recovery (i.e., 1SD below mean = − 1.70; β = 1.34, SE = 1.83, p > .10), CR was unrelated to externalizing symptoms. Cortisol reactivity to the acute stressor was not a significant moderator and there was no moderation for any stress-system variable on the link between cumulative risk and child internalizing symptoms.⁶ Results of all regression analyses are presented in Table 4.

Fig. 3.

Interactions between the number of risk factors and cortisol recovery on children’s externalizing behaviours

Note: Number of risk factors and stress system reactivity are shown and ± 1 SD from the mean for visualization of the reported moderation effects.

Table 4.

Results of regression analyses.

		Model 1
		Internalizing problems						Externalizing problems
				CI						CI
Variables	n	b(se)	t	LL	UL	Overall F	R2	b(se)	t	LL	UL	Overall F	R2
PEP	40
CR		0.70(1.27)	0.55	−1.88	3.28			1.14(1.13)	1.01	−1.15	3.43
PEP		0.19(0.16)	1.21	−0.13	0.52			. 19(0.16)	1.21	−0.13	0.52
CR x PEP		0.22(0.13)	1.58	−0.06	0.49	1.51	0.11	0.13(0.12)	1.07	−0.12	3.8	1.85	0.13
RSA	43
CR		1.07(1.33)	0.95	−1.22	3.36			1.03(0.93)	1.11	−0.85	2.90
RSA		4.25(1.65)	2.58*	0.91	7.58			4.25(1.65)	2.58**	0.91	7.58
CR x RSA		0.15(1.48)	0.10	−2.83	3.14	0.90	0.06	2.72(1.21)	2.25**	0.28	5.17	5.12	0.28
Cort reactivity	49
CR		1.58(1.01)	0.13	−0.46	3.61			1.20(0.93)	0.03*	0.12	3.87
Cort Reactivity		0.65(1.19)	0.54	−1.74	3.04			−0.82(1.09)	−0.75	−3.02	1.37
CR x Creactivity		0.55(1.10)	0.50	−1.67	2.77	0.82	0.05	0.27(1.01)	0.27	−1.77	2.30	0.07	0.13
Cort recovery	49
CR		1.54(1.03)	1.50	−0.53	3.61			1.41(0.92)	0.06	−0.44	3.27
Cort Recovery		−0.73(1.29)	−0.60	−3.32	1.86			2.23(1.15)	1.93*	−0.09	4.55
CR x Crecovery		0.70(1.18)	0.60	−1.67	3.07	1.21	0.07	2.12(1.05)	2.01*	−0.003	4.24	4.05	0.20

^*Significant at the 0.01 level.

^**Significant at the 0.05 level.

4. Discussion

The goal of this study was to characterize the links between CR and internalizing and externalizing symptoms in a sample of preschool-age children with exposure to a range of stressors. The potential moderating role of the HPAA and ANS systems was of particular interest to help advance our current understanding of how different stress physiological systems respond to environmental demands and how this dynamic responding impacts concurrent or subsequent behaviour. Particularly in individuals with repeated exposure to stress, physiological systems can experience strain (i.e., increased allostatic load) leading to individual differences in susceptibility to stress and an increased risk of psychopathology (McEwen and Stellar, 1993; Busso et al., 2017). CR seems to be particularly important for understanding the long-term consequences of repeated exposure to stress in light of research showing that multiple risk factors can induce excessive activation of stress systems (Shonkoff et al., 2012). Given that high CR is likely to be closely linked with chronic stress exposure, examining CR in the context of acute stress will help develop our understanding of stress management and the developmental trajectories of psychopathology related to stress. This study specifically addressed how stress exposure in early childhood (indexed by CR) predicts the early emergence of mental health symptoms in young children, and how individual differences in physiological responding moderates this link.

Results supported previous research that early environmental risk predicts early externalizing behaviours (Trentacosta et al., 2008; Ashford et al., 2008). Specifically, children with a greater index of CR had more externalizing symptoms than those with a low index of CR, which is consistent with early CR models showing that as the number of risk factors increase, the higher prevalence of developmental and clinical problems (Rutter, 1979). This model has been demonstrated across multiple studies examining the role of CR on children’s behavioural outcomes, with robust findings showing that children evidence more behaviour problems as they experience more risk factors (Appleyard et al., 2005; Atzaba-Poria et al., 2004; Atkinson et al., 2015). However, we found no association between CR and children’s internalizing symptoms, but this is supported by other work showing that CR more robustly predicts internalizing behaviours later in development than in early childhood (Ashford et al., 2008). Previous studies have found an association between CR and internalizing symptoms (Appleyard et al., 2005), but these findings are predominantly with children over the age of seven. Our results demonstrate that the effects of CR are still evident even in early childhood but that CR seems to be more relevant to externalizing problems in this 4–6 year old sample. Future studies should continue to explore this distinction using longitudinal methods to better understand independent developmental trajectories of both internalizing and externalizing symptoms.

There is increasing interest in how early adversity shapes and interacts with neurophysiological mechanisms to impact mental health development. Individual differences in stress system regulation may help explain the link between cumulative risk factors and the early emergence of mental health symptoms (Belsky and Pluess, 2009). Our findings are largely consistent with a diathesis stress model suggesting that, in the context of high CR, less PNS reactivity and a lack of cortisol recovery is associated with greater externalizing symptoms in children. This is in line with previous work that has shown low stress reactivity (i.e., less RSA withdrawal) is associated with externalizing symptoms in children experiencing greater adversity (Calkins et al., 2007; Obradović et al., 2010; Boyce et al., 2001) and contribute to a growing body of research suggesting varied associations between stress system function and behaviour in stressful developmental contexts (Obradović et al., 2010; Shonkoff et al., 2012; Wadsworth et al., 2019).

We also found that PNS reactivity was related to children’s externalizing problems at both moderate and high levels of CR, which largely supports a CR framework that as the number of risk factors increase, so do the number of negative outcomes (Appleyard et al., 2005; Atzaba-Poria et al., 2004; Atkinson et al., 2015; Rutter, 1979). Though reactivity was not significantly related to externalizing behaviours at low cumulative risk, visualization of the interaction effects may reveal important information to guide future research (Figs. 2a, 3a); greater RSA reactivity may serve a protective function in the context of high cumulative risk that is associated with better externalizing mental health outcomes, comparable to individuals with lower CR. It is also possible that a common underlying protective factor is linked to both PNS flexibility and low externalizing problems. Higher executive function and prefrontal cortex ‘top-down’ control has been theoretically and empirically linked to both greater PNS flexibility and lower externalizing problems (Kahle et al., 2018). Evolutionary-developmental models explaining individual differences in stress system function support the similar notion that children may develop alternative patterns of reactivity and associated behaviour to thrive, or survive, in different contexts, reflecting adaptive variation (Del Giudice et al., 2011). Although numerous other models exist to explain the moderating role of stress reactivity in the link between early adverse environments and children’s outcomes, it is also important to consider that early experiences shape the development of stress systems which consequently influences profiles of reactivity in other contexts. Future research might consider both mediating and moderating pathways by which physiological reactivity exerts effects on children’s outcomes in the context of early adversity. It is important to note here that follow-up analysis using residual scores produced slight differences in results when considering RSA reactivity such that CR was only related to externalizing problems in individuals with high reactivity, but not average reactivity. Residual scores take into account differences in baseline levels across the sample and define reactivity as the relative deviation from the predicted value. While residual scores have the advantage of limiting any unreliability of difference scores that do not take into account sample differences in baseline values, researchers have noted that because RSA is typically measured with good precision over short intervals, raw change may be preferred over residualized change and is more frequently used to compute RSA reactivity (Beauchaine et al., 2019).

When considering HPA activity, we found no moderating role of cortisol reactivity in the link between CR and internalizing and externalizing symptoms, which is in contrast to other work examining young children’s HPA activity in the context of repeated stress exposure. While several studies have found a moderating role of cortisol reactivity, findings are notably inconsistent; higher cortisol reactivity has been associated with greater externalizing symptoms in young children in adverse developmental contexts such as high parental hostility (Barrios et al., 2017), family stress (Obradović et al., 2010), and frequent maternal punishment (Hastings et al., 2011) while others have shown that blunted cortisol is related to externalizing behaviours and internalizing behaviours (e.g., Alink et al., 2008). It is also worth noting that there are differences regarding absolute thresholds of reactivity, and inconsistencies in analysis (e.g., using difference scores vs. residual scores) will produce slight variations in results. While we found no significant interaction between CR and cortisol reactivity using difference scores, the interaction was marginally significant using residualized scores. In theory, higher baseline cortisol levels may result in a lower capacity for change, which may not be accounted for using difference scores. In addition, the role of cortisol reactivity is relatively complex due to variations in activation related to repeated exposure to stress; there is an initial hyperactivation linked to chronic stress and a subsequent downregulation following high allostatic load (Danese and McEwen, 2012). It is also important to consider that cortisol reactivity is difficult to isolate in laboratory conditions, particularly in young children that show elevated cortisol related to entering a novel and unpredictable environment (Kryski et al., 2011). Researchers have noted that only a small minority of studies with young children report a mean increase in cortisol following a stressor paradigm (Gunnar et al., 2009); it is possible that some children showed consistently elevated cortisol from baseline so that reactivity profiles specific to the stressor task were not detected. Given a multitude of factors that affect HPA reactivity to acute stress under laboratory conditions, there has been increased interest in examining cortisol recovery as a more reliable and accurate measure of HPA flexibility, particularly in response to both a stressor challenge and the overall lab experience.

Regarding cortisol recovery, we showed that less cortisol recovery was related to more externalizing symptoms, specifically for children experiencing high cumulative risk. Despite a small number of studies examining cortisol recovery, our results extend findings that blunted or delayed cortisol recovery predicts greater symptoms of mental health problems (Nederhof et al., 2015; Schoorl et al., 2017). Similarly, it has been demonstrated that stressor paradigms characterized by uncontrollability and/or social-evaluative threat, as in the matching task employed in this study, are associated with greater delays in recovery (Dickerson and Kemeny, 2004). This has been observed in other studies employing similar tasks, with some finding a similar pattern of results to ours; for example, prolonged cortisol recovery after a perceived uncontrollable and socially evaluative task (e.g., Trier Social Stress Task) predicted greater psychopathology whereas cortisol increases associated with the task did not (e.g., Nederhof et al., 2015). These findings also fit well with work highlighting the effects of the early environmental factors, which show that a positive and supportive familial environment may buffer children’s stress response to have faster rates of recovery following an acute stressor (Gunnar and Hostinar, 2015), while the opposite is true in harsher family contexts. Our results add to a growing body of literature highlighting the importance of examining cortisol recovery for understanding risk for psychopathology.

Methodological characteristics are noted as limitations of the present study. First, with a small sample size of 54, the results should be interpreted carefully and replicated in future research. Future studies should draw a larger sample that may capture greater variability in cumulative risk profiles, stress system functioning, and associated symptoms. More specifically, the small sample size limited our ability to examine other important relationships among stress physiology and psychopathology that warrant attention, particularly how the interaction of stress systems contributes to risk for maladaptive behavioural outcomes and the implications of stress reactivity differences among individuals with greater cumulative risk. Future studies on the role of physiological regulation in predicting children’s mental health outcomes in the context of adverse family contexts should examine patterns of reactivity across multiple systems in order to more accurately capture possible biological vulnerability markers that confer risk for maladaptation. In addition, although we did not identify age, gender, or race as covariates, there is research suggesting they may contribute to differences in how biological factors contribute to increased risk for maladaptation in the context of adversity (Badanes et al., 2011; Kryski et al., 2011; Rudd et al., 2017). This should be explored in a larger sample to clarify the groups that are most at risk.

Another important avenue that was not addressed in this study is how autonomic recovery predicts children’s maladaptive outcomes in adverse conditions. Autonomic recovery could not be feasibly measured due to the methodological design of the larger study from which this sample was drawn (Roos et al., 2017). Future studies should examine both reactivity to and recovery from acute stress across multiple stress systems to differentiate their effects on children’s internalizing and externalizing problems. In terms of cumulative risk, this study relied on parents’ accurate reporting of their child’s early environment and did not include objective or alternate reports of household risk (e.g., noise, housing problems; Evans et al., 2007). It should also be noted that our sample on average did not have clinical levels of either internalizing or externalizing symptoms, so the results may not be directly applicable to populations selected to have clinical-levels of mental health impairment. However, given that symptoms operate on a continuum and that early emergence of symptoms may predict later clinical symptoms (Atkinson et al., 2015; Gutman et al., 2019), our findings may be key to understanding the window of opportunity for early prevention and intervention strategies to improve the course of psychopathological development.

Given the long-lasting effects that early adversity can have on adaptive functioning and mental health, there is a need to understand the underlying mechanisms that buffer or confer risk for maladaptive outcomes. Future research should seek to build this understanding using concurrent assessment of other biological assessment (e.g., neural activation, sympathetic nervous system) in the context of varying stressors. Examining ANS reactivity and recovery would also be of value, but this study did not include an appropriate post-stress recovery task to assess ANS recovery. It would also be informative for future studies to examine modifiable parenting factors that mediate effects of high CR. For example, there is considerable research implicating other factors such as harsh parenting (Bayer et al., 2011) and maternal mental disorder (Cents et al., 2013) as potential risk factors for early emerging problem behaviours in young children.

Taken together, our results are consistent with previous findings that exposure to cumulative stress in early childhood is associated with early emergence of mental health symptoms in young children, particularly in the context of multiple adversity exposure. We extended these findings by demonstrating the relevance of stress-system reactivity to mental health for children such that PNS and HPAA recovery is predictive of greater externalizing symptoms, particularly in children who have greater exposure to potential stressors. These findings suggest that problem behaviours in children associated with early adversity may be influenced by children’s ability to flexibly engage and regulate their PNS and HPAA, providing greater specificity in understanding why some children with repeated exposed to stress may be particularly vulnerable to psychopathology. These findings have the ability to inform interventions that enhance stress-management abilities as a way to promote resilient trajectories for mental and behavioural health in adversity-exposed children and their caregivers.