Childhood violence exposure and externalizing behaviors: a systematic review of the role of physiological biomarkers

Abstract

Childhood exposure to violence has been consistently linked to externalizing behaviors like delinquency and aggression. Growing evidence indicates that physiological biomarkers from the parasympathetic and sympathetic nervous systems (PNS and SNS) and hypothalamic-pituitary-adrenal (HPA) axis may moderate or mediate the relation between childhood violence exposure and externalizing behaviors. We conducted a systematic review to synthesize recent findings on physiological biomarkers as mediators and/or moderators of this association across the life course, using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Our search yielded 3,878 articles, of which 44 met inclusion criteria (describing a total of 46 independent studies). We found consistent evidence for blunted HPA-axis reactivity as a mediator of the relation between childhood violence exposure and subsequent externalizing behaviors, and for non-reciprocal PNS/SNS activation as moderators exacerbating this relation. However, the results of the majority of included studies that demonstrated significant moderating effects of physiological biomarkers varied by participant sex, type of childhood violence exposure, and type of stimuli used to induce physiological reactivity. The observed mixed findings are consistent with some theories that emphasize that both high and low stress reactivity can be adaptive depending on one’s early environment. These findings highlight the need for systematic explorations of heterogeneity, theory-driven research questions, and longitudinal studies that span multiple developmental periods and multiple biological systems. Clinical implications include the need to assess physiological biomarkers in treatment and intervention studies and the potential to target interventions based on both autonomic functioning and environmental contexts.

1. Introduction

Externalizing behaviors, including aggression and delinquency, are common among children and are the main reason for referral to child mental health services (Mazzucchelli and Sanders, 2018). Individuals with externalizing symptoms face academic, socialization, employment, and legal problems (e.g., juvenile delinquency and adult crime), and are at high risk for developing mental health disorders (e.g., oppositional defiant disorder, conduct disorder, antisocial personality disorder, substance use disorder) (Brook et al., 2011; Mazzucchelli and Sanders, 2018). Given the personal and community costs of externalizing behaviors, identifying potential mechanisms and influencing factors contributing to these behaviors may help inform more effective prevention, diagnosis, and intervention strategies.

1.1. Childhood violence exposure and externalizing behavior

Childhood violence exposure in different contexts (e.g., home, school, neighborhood), both direct (e.g., child abuse) and indirect (e.g., witnessing domestic violence), has consistently emerged as a strong predictor of later externalizing behavior (e.g., Evans et al., 2008; Fleckman et al., 2016; Gilbert et al., 2009). Up to 30% of children experience some form of physical or sexual abuse and up to 27% witness domestic violence (Finkelhor et al., 2009; Gilbert et al., 2009). Furthermore, children’s exposure to one type of violence increases the risk of exposure to other types (Finkelhor et al., 2007).

Several theoretical models have been developed to explain the role of biological mechanisms in the relation between childhood environmental stressors, like exposure to violence, and adverse developmental outcomes. Allostatic load theory posits that cumulative and repeated stressors may lead to a change in the functioning of stress response systems in ways that impede health (McEwen and Stellar, 1993). For instance, chronically high levels of the stress hormone cortisol damage hippocampal and cortical neurons. Thus, a downregulation of the stress system may be an allostatic response that is protective in the short term, but has longer term negative consequences for health and behavior (Lee et al., 2015). The allostatic load theory represents a mediation model, such that childhood violence exposure may lead to alterations in the physiological stress systems, which in turn increase the risk of adverse outcomes later in life (Atkinson et al., 2021; Finlay et al., 2022; Scheuer et al., 2018), including externalizing behaviors (e.g., Davies et al., 2007; Shenk et al., 2010; Sturge-Apple et al., 2012; Timmons et al., 2019; Timothy et al., 2019; White et al., 2017).

In addition, several theoretical vulnerability models characterize individual differences in sensitivity to environmental influences (Pluess, 2015), each with a distinct emphasis on sensitivity to negative (e.g., diathesis-stress), positive (e.g., vantage sensitivity), and both environmental qualities (e.g., differential susceptibility, Biological Sensitivity to Context [BSC] theory, Adaptive Calibration Model [ACM]). The diathesis-stress model (Monroe and Simons, 1991) asserts that certain biological characteristics render individuals more vulnerable to the impact of negative or stressful environments. In contrast, the vantage sensitivity model emphasizes that individuals with certain characteristics may benefit disproportionately from positive, enriched environments (Pluess and Belsky, 2013). Finally, the differential susceptibility model (Belsky et al., 2007) suggests that these characteristics may make individuals more susceptible to both positive and negative environmental influences. Two theoretical models have expanded on the differential susceptibility model by incorporating the role of stress response systems. Specifically, the BSC theory suggests that individuals with high stress reactivity are more vulnerable to negative environments, but also benefit more from positive environments (Boyce and Ellis, 2005). For example, high physiological reactivity has been associated with a wide variety of both stressors (e.g., aggression, adverse childhood events, parental depression, family conflict) and positive environments (e.g., parental warmth, supportive intervention) (Bush and Boyce, 2014). Moreover, the ACM emphasizes that patterns of stress reactivity across multiple stress response systems match the local environments during development to maximize one’s chances of survival and reproduction (Del Giudice et al., 2011). Changes in stress responsivity are considered adaptive. Under severe stress, high stress reactivity allows for threat detection and defense, while low reactivity promotes insensitivity to threat (Del Giudice et al., 2011). Indeed, in some studies, high and low stress responsivity patterns among children were found under both high and low stress family conditions (Berry et al., 2017; Del Giudice et al., 2012; Ellis et al., 2017). In general, these vulnerability models are examples of moderation models, positing that childhood violence exposure and aspects of the stress response system interact to increase the risk of adverse outcomes. Importantly, several studies of externalizing behavior offer empirical support to theories of differential susceptiability, BSC, and ACM (e.g., Bubier et al., 2009; El-Sheikh et al., 2011, 2009; Erath et al., 2011; Gordis et al., 2010; Kuhlman et al., 2018; McLaughlin et al., 2014; Obradović et al., 2010).

However, limited biopsychosocial models in the literature have focused specifically on externalizing behaviors. In particular, Raine (2002) proposed the biosocial model of violence, which posits that the interaction between environmental and biological risk factors exponentially increases the risk of violent behaviors. Expanding on Raine’s model, Liu’s biosocial model of childhood externalizing behavior (2004) highlighted potential mediating processes that may explain the relation between environmental and biological risk factors and externalizing behaviors, and moderating processes that may exacerbate those relations. While acknowledging mediating and moderating processes underlying the biopsychosocial effects on externalizing behaviors, Liu’s model primarily focuses on the influences of familial pathophysiological factors (e.g., maternal nutrition, illness or substance use during pregnancy, birth complications) and environmental factors during the pre- and perinatal periods in the development of childhood externalizing behaviors. Given the occurrence of externalizing behaviors through several developmental phases (i.e., from infancy through adolescence and beyond), a broader life course approach extending from childhood to young adulthood when considering externalizing behaviors may be helpful.

1.2. Key physiological biomarkers, violence exposure, and externalizing behavior

Growing evidence indicates that physiological biomarkers interact with the psychosocial environment to increase the risk for externalizing behaviors or are on the causal pathway between childhood violence exposures and externalizing behavior (Alink et al., 2012; Busso et al., 2017; Doom et al., 2020; Obradović et al., 2010). Biomarkers in both the autonomic nervous system (ANS) and neuroendocrine system are of particular interest, given their roles in stress response (Bauer et al., 2002). These systems also interact with each other to regulate homeostasis and allocate resources in the face of stress, highlighting the importance of a multi-system approach.

Physiological activity can be measured under different conditions, including during a resting state and in response to a stimulus. Specifically, physiological reactivity (i.e., the difference between resting levels and levels under stressful conditions) can be assessed in response to a variety of stressful stimuli (e.g., cognitive challenges, psychosocial stress tasks). Measuring physiological activity under both resting and reactive conditions may help inform underlying neural processes implicated in adaptive and pathological functioning (Zisner and Beauchaine, 2016). These physiological indices are widely regarded as markers of early adversity exposure and susceptibility to environmental influences but also have different implications for health and developmental outcomes (Obradović, 2012; Zisner and Beauchaine, 2016). In particular, resting physiological activity is commonly used as a marker of general arousal in the absence of stress, whereas reactive physiological activity is typically used as an indicator of active physiological coping in response or anticipation of stressors and challenges (Zisner and Beauchaine, 2016). Notably, choosing appropriate stimulus conditions that are relevant to key research questions is important, as different conditions may evoke different underlying physiological states (e.g., emotion dysregulation, affective responses, responsivity to social threat) that may influence the interpretation of findings (Zisner and Beauchaine, 2016).

The parasympathetic nervous system (PNS) is one branch of the ANS that regulates the “rest and digest” functions (McCorry, 2007) and helps to control homeostasis by regulating cardiac activity. Biomarkers of PNS activity include respiratory sinus arrhythmia (RSA) and cardiac vagal index (CVI). RSA (also known as high-frequency heart-rate variability) is the fluctuation in heart rate accompanying inspiration and expiration, which is typically captured through time domain peak-valley analysis, spectral analysis, or application of a band-pass filter (Beauchaine, 2001; Grossman and Taylor, 2007). High resting levels of RSA activity, indicating higher levels of PNS activity, are associated with slower heart rate. In response to stress, moderate decreases in RSA reactivity (i.e., vagal withdrawal) help to promote efficient arousal and increase metabolic output (Porges, 2007). Several studies have offered evidence for associations between violence exposure, lower resting RSA, and less RSA suppression in response to cognitive challenges or socioemotional stressors (see El-Sheikh and Erath, 2011). Lower levels of resting RSA and less RSA suppression in response to cognitive challenges or socioemotional stressors may be associated with increased risk of externalizing behaviors (Calkins et al., 2007; Graziano and Derefinko, 2013). CVI is derived from heart rate electrocardiography and has been found to be highly correlated with RSA (Allen et al., 2007).

The sympathetic nervous system (SNS) is another branch of the ANS, colloquially known as the “fight-or-flight” system (McCorry, 2007). In response to stress, SNS activity promotes increased heart rate and oxygen flow (McCorry, 2007). Some common, noninvasive ways to measure SNS activity include skin conductance level (SCL), cardiac pre-ejection period (PEP), cardiac sympathetic index (CSI), and salivary alpha-amylase (sAA). SCL refers to the measurement of electric conductivity of the skin in response to excitation of the sweat glands (McCorry, 2007). Higher resting SCL and SCL in response to conflict discussions have been identified as related to violence exposure (see El-Sheikh and Erath, 2011). Most studies have found that low resting SCL and low SCL reactivity in response to socioemotional tasks are associated with externalizing behaviors (Lorber, 2004), though others have implicated higher SCL reactivity in response to socioemotional tasks in externalizing behaviors (Beauchaine et al., 2008; El-Sheikh et al., 2007).

PEP is defined as the time between the onset of left ventricular depolarization and the ejection of blood into the aorta (Sherwood et al., 1990). Shorter PEPs are associated with SNS activation. Shorter resting PEP and lower PEP in response to social stress tasks are correlated with exposure to violence (Busso et al., 2017; Romero-Martínez et al., 2014), whereas longer resting PEP and higher PEP reactivity in response to cognitive challenges are correlated with more externalizing behaviors (Beauchaine et al., 2013). CSI is an index of heartbeat interval induced through the use of pharmacological blockades (Toichi et al., 1997) and has been found to be negatively correlated with aggression (Sturge-Apple et al., 2012). Lastly, sAA is a digestive enzyme released from the parotid gland and is associated with the adrenergic component of the SNS (Bosch et al., 2003), peaking about 5-10 minutes post-stressor and then quickly declining (Bauer et al., 2002; Gordis et al., 2008). Although some have regarded sAA as a promising SNS marker (Nater and Rohleder, 2009; Rohleder and Nater, 2009), others have argued that sAA may also be influenced by PNS (Bosch et al., 2003). While some studies indicate that higher sAA in response to social stress tasks is associated with exposure to violence (Gordis et al., 2010b; Rudolph et al., 2010), preliminary findings suggest that both low and high resting sAA are associated with more externalizing behavior (Keller and El-Sheikh, 2009).

Stressors activate the cascade of neuroendocrine systems, such as the hypothalamic-pituitary-adrenal (HPA) axis and hypothalamic-pituitary-gonadal (HPG) axis, which foster the release of cortisol and dehydroepiandrosterone-sulphate (DHEA-S). Cortisol, a by-product of the HPA axis cascade, is referred to as the stress hormone, as it fluctuates in response to environmental changes. Diurnal basal cortisol regulates basal bodily functioning, with basal levels of cortisol generally following a consistent circadian pattern. Following a stressor, cortisol levels typically take 10-15 minutes to rise (cortisol reactivity), followed by a decline in levels once the stressor ends (Dickerson and Kemeny, 2004). Several systematic reviews and meta-analyses have found that lower cortisol in response to social stress tasks and conflict discussions is associated with childhood adversity, including exposure to violence (Brindle et al., 2022; Bunea et al., 2017). In addition, meta-analytic evidence has revealed a small effect size for an increased risk of externalizing behaviors associated with lower basal cortisol levels (Alink et al., 2008). However, both lower and higher cortisol in response to cognitive challenges, social stress tasks, and conflict discussions may be associated with externalizing behaviors (Alink et al., 2008). In addition, DHEA-S is a sulphated derivative androgenic steroid hormone that plays a protective role in physical and mental health outcomes due to its glucocorticoid effects (Maninger et al., 2009). Exposure to violence has been preliminarily found to be linked with higher DHEA-S levels (Goodyer et al., 2001; Pico-Alfonso et al., 2004). Some studies have found positive relations between DHEA-S and externalizing behaviors, though results may vary by sex (Mundy et al., 2015; van Goozen et al., 2000). In addition, copeptin, a vasopressin precursor hormone, is a reliable and validated marker of HPA axis activation and severe stress (Katan et al., 2008). Whereas few studies have examined the link between copeptin and aggression, copeptin has been found to be an endocrine marker embedding childhood adversity (Coelho et al., 2016; Soares et al., 2021).

Examining activity across multiple systems concurrently may provide a more comprehensive understanding of how these systems work together to influence the development of emotional and behavioral problems (Bauer et al., 2002). The PNS and SNS play opposing roles in influencing autonomic functioning. Therefore, reciprocal coordination is considered adaptive in response to changes in the environment, whereas non-reciprocal coordination (i.e., co-inhibition or co-activation) is considered non-adaptive, as these systems may dampen, or cancel each other’s effects. Accumulating evidence has suggested that reciprocal coordination (i.e., high PNS and low SNS activity, or vice versa) may buffer the effects of childhood family stress on emotional and behavioral problems, whereas co-inhibition (i.e., low PNS and SNS activity) and co-activation (i.e., high PNS and SNS activity) may exacerbate the effects (El-Sheikh et al., 2013; Gordis et al., 2010a; Philbrook et al., 2018).

1.3. Past reviews of childhood violence exposures, physiological biomarkers, and externalizing behavior

Whereas past reviews have demonstrated that childhood violence exposure has long-term negative consequences on autonomic, neuroendocrine, genetic, and brain functioning (e.g., Bernard et al., 2017; Bunea et al., 2017; Ehlert, 2013; Finlay et al., 2022; Khoury et al., 2019; McCrory et al., 2011; Moffitt and Klaus-Grawe 2012 Think Tank, 2013), these reviews of biological effects of violence have largely not explored the impacts on behavioral outcomes. Furthermore, past reviews have emphasized the mediating effects, but not moderating effects, of biological mechanisms in the relation between childhood adversity and developmental outcomes. A meta-analysis and systematic review have explored moderating roles of different genotypes (e.g., MAOA, DRD4, COMT gene) in the relation between childhood violence exposure and externalizing and antisocial behavior (Byrd and Manuck, 2014; Weeland et al., 2015), but were restricted to DNA/genes as the biological mechanisms of interest. Currently, only one review attempted to summarize the moderating and mediating roles of physiological biomarkers in the association between early life adversity and later outcomes, focusing on the effect of child maltreatment on ANS functioning and psychopathology in childhood (Young-Southward et al., 2020). Although these authors found a consistent pattern of blunted resting and reactive SNS in response to cognitive challenges, social stress tasks, and conflict discussions among maltreated children, they observed mixed findings for the effects of maltreatment on PNS functioning. The moderating or mediating role of ANS functioning in the association between childhood maltreatment and psychopathology risk was also inconsistent. Focusing on a more homogeneous outcome (e.g., externalizing behaviors) and including studies that incorporate a multi-system approach may help provide a clearer picture of the role of physiological factors in the relation between childhood violence exposure and externalizing behaviors over the life course.

1.4. Present systematic review

This review aims to synthesize recent evidence on physiological biomarkers from the PNS, SNS, and neuroendocrine systems as mediators and/or moderators of the association between childhood violence exposure and externalizing behavior, and to extend prior theoretical work that explicates the relations among these factors. We expect this review to highlight areas for potential clinical and population-level intervention and provide insight into future research directions.

2. Materials and methods

We conducted this systematic review according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al., 2009). Our inclusion and exclusion criteria, search strategy, and review process were specified in advance and documented in a systematic review protocol (see Appendix).

2.1. Inclusion and exclusion criteria

We used the following inclusion criteria: (1) peer-reviewed English language papers reporting empiric/quantitative results, published since 2005; (2) studies conducted among humans; and (3) studies that included one or more physiological biomarkers of PNS, SNS, or neuroendocrine activity (e.g., RSA, SCL, cortisol) as a mediator or moderator of the relation between some measure of childhood exposure to violence and some measure of externalizing behavior in childhood, adolescence, or adulthood. Specifically, we included measures associated primarily with a single activation system (PNS, SNS, or neuroendocrine system), though acknowledging that most biomarkers are at least somewhat influenced by multiple systems. We only included studies published since 2005 in order to focus on the most recent evidence, given constant progress in measuring and understanding physiological biomarkers (Zisner and Beauchaine, 2016). Studies that included retrospective self-reports of childhood violence exposures (from an adult sample) or data prospectively collected during childhood were included, as were studies that examined either basal or reactive physiological biomarkers. Given heterogeneity in the tasks used to elicit a reactive psychophysiological response, we included studies that assessed physiological reactivity using different types of stress tasks, broadly classified as cognitive challenges (e.g., star-tracing task), psychosocial evaluative tasks (e.g., Trier Social Stress Test), socioemotional tasks (e.g., viewing or listening to emotional content), and physical discomfort tasks (e.g., cold pressor task) (El-Sheikh and Erath, 2011; National Advisory Mental Health Council Workgroup on Tasks and Measures for Research Domain Criteria (RDoC), 2016). We also included studies that examined mediating pathways without formally testing a full mediation model so as not to bias our findings towards studies that found statistically significant results; for example, some studies did not proceed with statistical tests of mediation if at least one part of the mediating pathway (e.g., the relation between the mediator and the outcome) was found to be non-significant.

We excluded studies not meeting the above inclusion criteria, including commentaries, editorials, review papers, meta-analyses, and animal studies. Furthermore, we excluded studies examining childhood adversity measures that did not focus on a specific violence component (e.g., poverty, childhood neglect, early life stress, composite measures of adversity). In addition, although bullying and sibling abuse are important sources of childhood violence that may have adverse effects on development (Wolke et al., 2015), we excluded studies focusing solely on bullying and sibling-related conflicts because they have common and normative influences on children’s socialization processes (Laursen and Pursell, 2009; Whiteman et al., 2009) and are heterogeneous experiences that are difficult to measure objectively (Kettrey and Emery, 2006). Typically, the construct of externalizing behavior consists of three components: aggressive, disruptive, and hyperactive behaviors (Hinshaw, 1987). We restricted our focus to aggression and delinquency because these behaviors are more likely to be associated with subsequent criminal and violent behavior (Liu, 2004). Thus, we additionally excluded studies assessing only symptoms of attention deficit hyperactivity disorder or other externalizing behavior-related constructs such as behavioral disinhibition or substance abuse. Lastly, we excluded studies that investigated associations between biological mechanisms and childhood violence exposure without externalizing behaviors as an outcome variable.

2.2. Search strategy

We searched for articles indexed on PubMed, Web of Science, and PsycINFO, using a comprehensive search strategy aimed at maximizing the sensitivity of our search for identifying relevant articles. We included search terms for: (i) physiological biomarkers (“autonomic functioning” OR “sympathetic nervous system” OR “parasympathetic nervous system” OR “skin conductance” OR “respiratory sinus arrhythmia” OR “heart rate” OR “pre-ejection period” OR “stress response functioning” OR “neuroendocrine regulation” OR “HPA axis” OR cortisol OR testosterone OR DHEA OR alpha-amylase OR “psychophysiological responding”), AND (ii) childhood violence exposure (“child* violence exposure” OR “child* maltreatment” OR “child* abuse” OR “child* victimization” OR “harsh parenting” OR “harsh discipline” OR “parental psychological control” OR “marital conflict” OR “intimate partner violence” OR “domestic violence” OR “community violence”), AND (iii) externalizing behaviors (aggression OR “aggressive behavior” OR “violence perpetration” OR “externalizing behavior” OR “antisocial traits” OR “antisocial personality disorder” OR callous-unemotional OR “conduct disorder” OR “oppositional defiant disorder”). To capture any citations we may have missed with our search strategy, we also conducted manual searches for additional articles, including a “backward” reference search of included articles, a “forward” reference search of papers citing included articles, and a search of reference lists of key literature reviews (Ehlert, 2013; Moffitt and Klaus-Grawe 2012 Think Tank, 2013).

2.3. Study selection and eligibility assessment

After we removed duplicates, two independent reviewers screened studies based on their title and abstract. Full-text articles were then reviewed by two independent reviewers for inclusion. The reviewers met to discuss and resolve any disagreements to reach consensus. Data were extracted from each included study using a standard extraction form, with the accuracy of extraction verified by at least one other author independently. Studies were grouped by biological system (i.e., PNS, SNS, neuroendocrine, and multi-system), biosocial model (mediation and moderation), and basal vs. reactive physiological activity for synthesis.

3. Results

3.1. Characteristics of selected studies

Our search yielded 3,878 records, with 3,650 remaining after duplicates were removed (Figure 1). After screening titles and abstracts, we reviewed 100 full-text articles for inclusion. We included a final total of 44 articles in this systematic review, which presented the results of 46 independent studies (one study included results from three separate studies) (El-Sheikh et al., 2009).

Figure 1.

Flow chart of study selection following PRISMA guidelines

^a One included article reported findings from three independent studies, for a total of 46 included studies

The included studies are summarized in Table 1. The majority of studies were conducted in the United States (n=39). Most of the included studies recruited male and female participants with a diverse racial/ethnic background, with the exception of three studies that recruited boys/men only (Cima et al., 2008; Peckins et al., 2018; Timothy et al., 2019), one that recruited girls/women only (Shenk et al., 2010), and studies conducted in countries of ethnic homogeneity (e.g., the Netherlands, India, China) (Cima et al., 2008; Peng et al., 2021; Timothy et al., 2019).

Table 1.

Overview of included studies^a

Authors (Year)	Participantsb	n	Sex	Race/ethnicityc	Study design	Mean aged	Other covariates	Biomarkers	Stimuli	Biosocial model
Alink et al. (2012)	Maltreated and non-maltreated children attending a week-long summer camp for inner-city children	236	128 boys 108 girls	53% Black 20% other 19% white 8% Latinx	Longitudinal (1 year; 2 waves)	W1: 7.64 yrs (SD = 1.36)	Age, sex	Salivary cortisol	n/a	Mediation
Bubier et al. (2009)	Children in three inner-city elementary schools	57	29 boys 28 girls	94% Black 6% Latinx	Longitudinal (11 months; 2 waves)	W1: 7.77 yrs (SD = 1.08)	Family income, sex	PEP, RSA	Socioemotional task and cognitive challenge	Moderation
Busso et al. (2017)	Adolescents recruited from schools, after-school programs, medical clinics, and the wider community	169	75 boys 94 girls	41% white 18% Black 18% Latinx 15% other 8% Asian	Cross-sectional	14.9 yrs (SD = 1.4)	Poverty	DHEA-S, PEP, RSA, salivary cortisol	Psychosocial evaluative task	Mediation
Chen et al. (2015)	Adolescents enrolled in the Philadelphia Healthy Brains and Behavior project	425	213 boys 212 girls	80% Black 12% white 8% other	Cross-sectional	11.87 yrs (SD = 0.60)	BMI, ethnicity, income, internalizing problems, pubertal stage, sex	sAA, salivary cortisol	n/a	Moderation
Cima et al. (2008)	Adult prison inmates and undergraduate controls in the Netherlands	74	74 men	100% Dutch	Cross-sectional	28.39 yrs (SD = 9.30)	Age	Salivary cortisol	n/a	Mediation
Coelho et al. (2016)	Subset of children in the High Risk Cohort Study for Psychiatric Disorder, recruited from 57 schools in 2 cities in Brazil	136	68 boys 68 girls	60% white 40% non-white	Cross-sectional	9.21 yrs (SD = 1.69)	Age, internalizing behaviors, race, sex	Serum co-peptin	n/a	Mediation
Cook et al. (2018)	Adolescents in 9th-11th grade who lived with a female parent/caregiver, recruited through community postings, a commercial mailing list, community outreach, and school partnerships in the Northeast	100	32 boys 68 girls	78% white 22% non-white	Longitudinal (1 year; 2 waves)	W1: 15.09yrs (SD = 0.98)	Baseline autonomy, baseline externalizing and internalizing problems, sex	Salivary cortisol	Psychosocial evaluative task	Moderation
Cui et al. (2019)	Adolescents recruited from disadvantaged communities in urban and rural areas in the southern Midwest	57	31 boys 26 girls	58% Black 21% white 18% other 4% Latinx	Cross-sectional	13.19 yrs (SD = 2.55)	Age, baseline RSA, changes in respiration rate, sex	RSA	Socioemotional task	Moderation
Davies et al. (2007)	Kindergarten children from a moderately-sized metropolitan area and a small city	178	79 boys 99 girls	77% white 16% Black 5% Latinx 1% Asian 1% other	Longitudinal (2 years; 2 waves)	W1: 6.0 yrs (SD = 0.48)	Baseline externalizing behaviors, parental warmth, race, SES, sex, time of cortisol measurement	Salivary cortisol	Socioemotional task	Mediation
DePasquale et al. (2019)	Maltreat and non-maltreated children attending a week-long summer camp for inner-city children	365	197 boys 168 girls	69% Black 20% white 11% other	Longitudinal, (1 year; 2 waves)	W1: 9.3 yrs (SD = 0.87)	Age, race, sex, stressful life events, time since waking	Salivary cortisol	n/a	Mediation
Doom et al. (2020)	Subset of adolescent participants from the Fragile Families and Child Wellbeing Study who underwent an in-person assessment for the Study of Adolescent Neural Development	171	69 boys 102 girls	76% Black 14% white 5% Latinx 5% other	Longitudinal (15 years; 5 waves)	W5: 15.3 yrs (SD = 0.4)	Age, BMI, hair-related measures, medication use, mother’s age at child’s birth, mother’s education, mother’s marital status, poverty at birth and age 15, race/ethnicity, sex	Hair cortisol	n/a	Mediation
El-Sheikh (2005)	Children from two-parent families recruited from the community through birth announcements and advertisements	180	89 boys 91 girls	66% white 27% Black 8% other	Cross-sectional	9.70 yrs (SD = 1.99)	Age, baseline SCL, observed and self-reported anger, fear, and sadness after the laboratory challenge, race, SES	SCL	Socioemotional task	Mediation and moderation
El-Sheikh & Whitson (2006)	Children in two-parent families recruited from the community through birth announcements and advertisements	180	89 boys 91 girls	66% white 27% Black 8% other	Longitudinal (2 years; 2 waves)	T1: 9.70 yrs (SD = 1.99)	Age, baseline RSA, race, SES	RSA	Socioemotional task	Moderation
El-Sheikh et al. (2007)	Children from two-parent families recruited from the community through birth announcements and advertisements	157	86 boys 71 girls	76% white 17% Black 7% other	Longitudinal (2 years; 2 waves)	W1: 9.31 yrs (SD = 1.97)	Age, baseline externalizing behaviors, baseline SCL, ethnicity, SES, sex	SCL	Socioemotional task and cognitive challenge	Mediation and moderation
El-Sheikh et al. (2009)	Public school students in two-parent households, enrolled in 3 independent studies	176	78 boys 98 girls	69% white 33% Black	Cross-sectional	8.69 yrs (SD = 0.40)	Age, BMI, ethnicity, family SES, sex	RSA, SCL	Socioemotional task and cognitive challenge	Moderation
		251	123 boys 128 girls	64% white 35% Black		8.23 yrs (SD = 0.73)	Age, ethnicity, family SES, sex
		150	75 boys 75 girls	67% white 27% Black 6% other		9.27 yrs (SD = 1.95)	Age, ethnicity, family SES, sex
El-Sheikh et al. (2011a)	Children from three school systems	251	123 boys 128 girls	64% white 36% Black	Longitudinal (3 years; 3 waves)	W1: 8.23 yrs (SD = 0.73)	BMI, initial ADH symptoms, race, SES	RSA, SCL	Cognitive challenge	Moderation
El-Sheikh & Hinnant (2011b)	Children from two independent community samples	413	194 boys 219 girls	67% white 33% Black	Longitudinal (3 years)	T1: 8.13 yrs (SD = 0.33)	Family income, race	RSA	Socioemotional task and cognitive challenge	Moderation
Erath et al. (2009)	Children from two-parent families in three school districts in a small town	251	123 boys, 128 girls	64% white 36% Black	Cross-sectional	8.23 yrs (SD = 0.73)	Age, child-reported internalizing symptoms, ethnicity, marital conflict, SES, sex	SCL	Socioemotional task and cognitive challenge	Moderation
Erath et al. (2011)	Children from two-parent families in three school districts in a small town	251	123 boys, 128 girls	64% white 36% Black	Longitudinal (3 years; 3 waves)	W1: 8.23 yrs (SD = 0.73)	Age, baseline child-reported internalizing symptoms, baseline marital conflict, ethnicity, SES, sex	SCL	Socioemotional task and cognitive challenge	Moderation
Gordis et al. (2010)	Maltreated youth from DCFS reports and non-maltreated comparison youth from similar neighborhoods	362	187 boys, 175 girls	39% Latinx 38% Black 13% other 11% white	Cross-sectional	12.1 yrs (SD = 1.21)	Age, ethnicity, income, parent education, sex	RSA, SCL	Socioemotional task	Moderation
Gowin et al. (2013)	Adults on parole or probation or who drank 2-12 alcoholic beverages per week, recruited through advertisements in targeted zip codes based on SES and education level	67	--	--	Cross-sectional	31.5 yrs (SD = 9.6)	n/a	Salivary cortisol	20mg dose of cortisol	Mediation
Hagan et al. (2014)	Young adults recruited from an undergraduate psychology course who scored in the highest or lowest quartiles of family conflict	88	44 men, 44 women	59% white 26% Latinx 7% Black 5% Asian 3% other	Cross-sectional	18.67 yrs (SD = 0.97)	Average weekly alcohol use, current family conflict, internalizing problems, smoking status	Salivary cortisol	Psychosocial evaluative task	Mediation and moderation
Hastings et al. (2011)	Children from 3 independent community studies (the Concordia Longitudinal Risk Project, the Daycare and Preschool Adjustment Study, and the Shame in Childhood Study) in Canada	402	206 boys 196 girls	83% white	Cross-sectional	4.01 yrs (SD = 0.71)	Age, inhibition, internalizing behaviors, sex, study, time of saliva collection, time of visit	Salivary cortisol	Socioemotional task	Moderation
Hinnant et al. (2015)	Children participating in the Family Stress and Youth Development: Bioregulatory Effects project, recruited from the community	251	122 boys, 129 girls	64% white 36% Black	Longitudinal (8 years; 4 waves)	W1: 8.23 yrs (SD = 0.72)	Age, change in harsh parenting, ethnicity, family income, marital conflict	RSA	Cognitive challenge	Moderation
Huffman et al. (2020)	Children recruited by community liaisons in a small city	101	43 boys, 58 girls	73% Black 13% white 8% other 6% Latinx	Longitudinal (1 year; 2 waves)	W1: 10.28yrs	Age, baseline aggression and delinquency, household income, sex	PEP, RSA	Psychosocial evaluative task	Moderation
Katz (2007)	Children recruited through advertisements in preschools, newspapers, and doctors’ offices	130	81 boys, 49 girls	89% white 6% Latinx 4% Black 2% other	Cross-sectional	5.04 yrs	n/a	RSA	Socioemotional task	Moderation
Koss et al. (2014)	Children in second grade	195	88 boys, 107 girls	77% white 17% Black 6% other	Longitudinal (6 years; 2 waves)	W1: 7.99 yrs (SD = 0.53)	Age, conflict stimulus, family income, pre-task cortisol and sAA, sex, time of saliva collection	sAA, salivary cortisol	Socioemotional task	Moderation
Kuhlman et al. (2018)	Adolescents recruited from the community via flyers, local advertisements, and referrals from clinicians	121	64 boys 57 girls	70% white 19% other 6% Black 3% Latinx 2% Asian	Cross-sectional	12.77 yrs (SD = 2.26)	Age, baseline cortisol, childhood non-intentional trauma, clinically significant internalizing symptoms, sex, total cortisol response to the task	Salivary cortisol	Psychosocial evaluative and physical discomfort task	Moderation
Lipscomb et al. (2018)	Children participating in Cohort I of the Early Growth and Development Study of adopted children and their birth and adoptive parents in 10 states	361	155 boys, 206 girls	58% white 22% other 11% Black 9% Latinx	Longitudinal (1.5 years; 2 waves)	W1: 4.62 yrs (SD = 1.89)	Adoptive parents’ education and income, age, sex	Salivary cortisol	n/a	Moderation
Madden & Shaffer (2019)	Young adult university students who were in a monogamous, heterosexual dating relationship of at least one month’s duration	57	11 men, 46 women	67% white 14% Asian 9% Latinx 8% other 2% Black	Cross-sectional	19.47 yrs (SD = 1.39)	Length of dating relationship, sex	Salivary cortisol	Psychosocial evaluative task	Moderation
McKernan & Lucas-Thompson (2018)	Adolescents from two-parent families in a moderately sized community	153	73 boys, 80 girls	49% white 27% other 17% Black 6% Asian 1% Latinx	Longitudinal (1 year; 2 waves)	W1: 12.92yrs (SD = 2.16)	Age, baseline externalizing behaviors, caffeine consumption, ethnicity, family SES, height, sex	RSA, SCL	Psychosocial evaluative task	Moderation
McLaughlin et al. (2014)	Adolescents recruited from schools, after-school programs, medical clinics, and the community	168	74 boys, 94 girls	41% white 18% Black 18% Latinx 15% other 8% Asian	Cross-sectional	14.9 yrs (SD = 1.36)	Age, respiration rate, sex	PEP, RSA	Psychosocial evaluative task	Moderation
Obradovic et al (2011)	Children with partnered caregiver recruited from kindergarten classrooms in 6 public schools	260	130 boys, 130 girls	49% white 24% other 12% Asian 11% Black 4% Latinx	Cross-sectional	5.33 yrs (SD = 0.31)	Sex	PEP, RSA	Socioemotional task and cognitive challenge	Moderation
Peckins et al. (2018)	Adolescents participating in the Pitt Mother & Child Project in Pittsburgh, Pennsylvania	190	190 boys	53% white 36% Black 11% other	Longitudinal (5 years; 3 waves)	W1: 12 yrs	Baseline antisocial behavior, marital conflict at younger ages, hostile attributions, neighborhood dangerousness, pro-violence attitudes, race	Salivary cortisol	Psychosocial evaluative task	Moderated mediation
Peng et al. (2021)	Adolescents recruited from one junior high school in China whose parents had been living together for at least 3 years	332	181 boys, 151 girls	95% Chinese Han ethnicity	Longitudinal (3 years; 3 waves)	W1: 13.7 yrs (SD = 0.8)	Baseline RSA, internalizing problems, sex	RSA	Psychosocial evaluative task	Moderation
Philbrook et al. (2018)	Adolescents participating in the Family Stress and Youth Development: Bioregulatory Effects project	252	119 boys, 133 girls	66% white 34% Black	Longitudinal (2 years; 3 waves)	W1: 15.79yrs (SD = 0.81)	Family SES at age 16, race/ethnicity, sex	RSA, SCL	Cognitive challenge	Moderation
Saxbe et al. (2012)	Children recruited through advertisements and referrals with parents who had lived together for at least 3 years	54	29 boys, 25 girls	35% Latinx 28% white 19% Black 11% Asian 7% other	Longitudinal (6 years; 4 waves)	W4: 15.2 yrs (SD = 0.8)	n/a	Salivary cortisol	Psychosocial evaluative task	Moderation
Shenk et al. (2010)	Maltreated children referred from CPS and non-maltreated comparison participants recruited via community advertisements	129	129 girls	54% white 42% Black 3% Latinx 1% Asian	Longitudinal (13 years; 3 waves)	W1: 11.06yrs (SD = 2.98)	Age, externalizing and internalizing behaviors in late adolescence, minority status, main effects for vagal withdrawal and cortisol AUC, vagal tone during relaxation	RSA, salivary cortisol	Cognitive challenge	Mediation
Sturge-Apple et al. (2012)	Toddlers recruited through agencies serving disadvantaged families in a moderately sized metropolitan area	201	109 boys, 92 girls	56% Black 23% white 11% Latinx 10% other	Longitudinal (1 year; 2 waves)	W1: 26 mo (SD = 1.69)	“Hawk” and “dove” latent profiles	CSI, CVI, salivary cortisol	n/a	Mediation
Tabachnick et al. (2020)	Children participating in a parenting intervention after CPS referral in infancy and a non-CPS-referred comparison group recruited in middle childhood	123	63 boys, 60 girls	60% Black 27% other 13% white	Longitudinal (9 years; 2 waves)	W2: 9.46 yrs (SD = 0.34)	n/a	RSA	Socioemotional task	Moderation
Timmons et al. (2019)	Young adults who were in a dating relationship for at least 2 months	218	106 men, 112 women	28% white 24% Latinx 20% other 16% Black 13% Asian	Cross-sectional	23.1 yrs (SD = 3.0)	n/a	SCL	Socioemotional task	Mediation
Timothy et al. (2019)	Children of treatment-seeking fathers with onset of alcohol dependence before age 25 and one or more first degree relatives with alcohol dependence, and healthy matched controls in India	100	100 boys	100% Kannada	Cross-sectional	11.25 yrs (SD = 2.03)	n/a	Salivary cortisol	Psychosocial evaluative task	Mediation
Vaillancourt et al. (2018)	Children from 41 classrooms within 19 high-quality child care centers in Canada	198	100 boys, 98 girls	84% white	Cross-sectional	2.80 yrs (SD = 0.42)	Age in months, classroom, hours per week in child care, household income	Salivary cortisol	n/a	Moderation
White et al. (2017)	Maltreated and non-maltreated children and adolescents participating in the AMIS project in Germany	537	265 boys, 272 girls	--	Cross-sectional	9.98 yrs (SD = 3.13)	Age, BMI, caregiver education, employment, and income, hair-related measures, non-abuse life adversities, pubertal status, recent stressful events	Hair cortisol	n/a	Mediation

^aAbbreviations: CPS – Child Protective Services; CSI – cardiac sympathetic index; CVI – cardiac vagal index; DHEA-S – dehydroepiandrosterone-sulphate; PEP – pre-ejection period; RSA – respiratory sinus arrhythmia; sAA – salivary alpha-amylase; SCL – skin conductance level; UK – United Kingdom

^bAll included studies were conducted in the United States, except 7 studies as noted in the participant description

^cMay not add up to 100% because of rounding

^dFor longitudinal studies, mean age at the specified wave

Of the 46 included studies, 24 used a cross-sectional analytic design, whereas 22 used a longitudinal design with two or more time points. The mean age of participants at the outcome assessment ranged from 37 months old to 31.5 years old, with 20 studies assessing outcomes in early or middle childhood (less than 11 years old), 20 in adolescence (11-17 years old), and 6 in adulthood (18 years and older). Specific outcomes of interest included externalizing behaviors (n=30), aggression (n=5), antisocial behavior (n=4), delinquency (n=2), disruptive/aggressive behaviors (n=2), interpersonal dating conflict (n=2), and conduct problems (n=1). The plurality of studies relied on multiple reporters when assessing outcomes (n=21), whereas 12 studies used only self-report, and other studies relied on reports of parents or primary caregivers (n=12) or teachers (n=1). Childhood violence exposures included domestic violence/marital conflict (n=16), child maltreatment (n=13), harsh parenting or parental psychological control (n=11), peer victimization (n=1), and combinations of multiple childhood violence exposures (n=5; e.g., child maltreatment and community violence).

Twenty-eight studies tested moderation effects of a physiological biomarker on the link between childhood exposure to violence and externalizing behaviors, whereas 14 studies assessed one or more physiological biomarkers as mediators, and four studies examined both moderation and mediation. Physiological biomarkers included cortisol (n=16), heart rate-related biomarkers (e.g., RSA, CVI, CSI; n=7), SCL (n=5), and copeptin (n=1). The remaining studies (n=17) measured more than one physiological biomarker (e.g., PEP and RSA, RSA and SCL, sAA and salivary cortisol). Overall, 11 studies assessed physiological biomarkers only at rest, whereas the remaining 35 studies assessed the reactivity of physiological biomarkers in response to stressors or biomarkers at both resting and reactive states. We summarize findings from the included studies below, by type of biosocial model and by system (mediation analyses in Table 2, moderation by PNS and SNS biomarkers in Table 3, moderation by neuroendocrine biomarkers in Table 4, and moderation by multi-system biomarkers in Table 5).

Table 2.

Summary of included studies examining biomarkers as mediators of the relation between childhood violence exposure and externalizing behaviors^a

First author (Year)	n	% male	Mean age	Study design	Childhood violence exposure	Biological system	Biomarker(s)	Stimuli	Outcome	Outcome measure	Major findingsb
Basal activity of biomarker as mediator of relation between childhood violence exposure and externalizing behavior outcome
Alink (2012)	236	54%	7.64y	L	Child abuse	Neuroendocrine	Salivary cortisol	n/a	Dis./agg. Behavior	Peer noms., PEI	n.s. for diurnal salivary cortisol
Cima (2008)	74	100%	28.4y	C	Child abuse	Neuroendocrine	Salivary cortisol	n/a	Aggression	BPAQ	n.s. for diurnal salivary cortisol
Coelho (2016)	136	50%	9.21y	C	Child abuse	Neuroendocrine	Serum co-peptin	n/a	EB	CBCL	n.s. for serum co-peptin
DePasquale (2019)	365	54%	9.3y	L	# child abuse subtypes	Neuroendocrine	Salivary cortisol	n/a	EB, aggression	CBCL-TRF, PEI	n.s. for morning cortisol levels
Doom (2020)	171	40%	15.3y	L	Harsh parenting	Neuroendocrine	Hair cortisol	n/a	AB	CBCL, SRD, K-SADS	n.s. for hair cortisol concentration
Sturge-Apple (2012)	201	54%	26m	L	Harsh parenting	PNS, SNS, Neuroendocrine	CVI, CSI, Salivary cortisol	n/a	EB	CBCL	n.s. for basal CVI Harsh parenting → elevated CSI → increased EB (in “hawk” children) n.s. for salivary cortisol
White (2017)	537	49%	9.98y	C	Child abuse	Neuroendocrine	Hair cortisol	n/a	EB	CBCL, SDQ	Child abuse → lower HCC → higher EB
Reactivity of biomarker as mediator of relation between childhood violence exposure and externalizing behavior outcome
Busso (2017)	169	44%	14.9y	C	Interpersonal violence	PNS, SNS, Neuroendocrine	RSA, PEP, DHEA-S and salivary cortisol	Psychosocial evaluative task	EB	CBCL-YSR	n.s. for reactive RSA n.s. for reactive PEP Interpersonal violence → blunted HPA-axis reactivity → higher EB
Davies (2007)	178	44%	6.0y	L	Marital conflict	Neuroendocrine	Salivary cortisol	Socioemotional task	EB	CBCL	Marital conflict → blunted cortisol reactivity → increased EB
El-Sheikh (2005)	180	49%	9.67y	C	Marital conflict	SNS	SCL	Socioemotional task	EB	PIC	n.s. for SCL-R
El-Sheikh (2007)	157	55%	9.31y	L	Marital conflict	SNS	SCL	Socioemotional task and cognitive challenge	EB	PIC	n.s. for SCL-R
Gowin (2013)	67	--	31.5y	C	Child abuse	Neuroendocrine	Salivary cortisol	20mg dose of cortisol	Aggression	BPAQ	n.s. for salivary cortisol reactivity
Hagan (2014)	88	50%	18.67y	C	Child abuse	Neuroendocrine	Salivary cortisol	Psychosocial evaluative task	EB	ASR	n.s. for salivary cortisol reactivity
Shenk (2010)	129	0%	11.06y	L	Child sexual abuse	Multi-system	RSA and salivary cortisol	Cognitive challenge	AB	ABF	Child sexual abuse → asymmetrical HPA-axis/PNS functioning → higher AB
Timmons (2019)	218	49%	23.1y	C	Harsh parenting	SNS	SCL	Socioemotional task	DAP	HDPTEO	Harsh parenting → higher SCL-R → DAP (in women)
Timothy (2019)	100	100%	11.25y	C	Interpersonal violence	Neuroendocrine	Salivary cortisol	Psychosocial evaluative task	EB	SDQ	Interpersonal violence → blunted cortisol reactivity → higher EB

^aAbbreviations: BPI – Berkeley Puppet Interview; C – cross-sectional; CBCL – Child Behavior Checklist; CBCL-YSR: Child Behavior Checklist-Youth Self-Report; CSI-4 – Child Symptom Inventory; CVI – cardiac vagal index; DISC – Diagnostic Interview Schedule for Children; EB – externalizing behaviors; HBQ – MacArthur Health and Behavior Questionnaire; L – longitudinal; PBFS: Problem Behavior Frequency Scale; PIC: Personality Inventory for Children; PIC2: Personality Inventory for Children-II; RPQ: Reactive-Proactive Questionnaire; R-PRPA – Revised Parent Rating Scale for Reactive and Proactive Aggression; RSA – respiratory sinus arrhythmia; SRASBM: Self-Report of Aggression and Social Behavior Measure

^bMajor findings indicate non-statistically significant (n.s.) estimated mediation effects, and statistically significant mediators of the relation between the exposure and outcome. Specific exposures and outcomes are specified if more than one was assessed in the study, and any population sub-groups to which significant effects were confined are listed in bold.

Table 3.

Summary of included studies examining parasympathetic and sympathetic nervous system biomarkers as moderators^a

First author (Year)	n	% male	Mean age	Study design	Childhood violence exposure	Biomarker(s)	Stimuli	Outcome	Outcome measure	Major findingsb
Basal activity of PNS biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Bubier (2009)	57	51%	7.77y	L	Harsh parenting	RSA	n/a	EB	CSI-4	n.s. for basal RSA
El-Sheikh (2011a)	251	49%	8.23y	L	Marital conflict	RSA	n/a	Delinquency	PIC2	+ for lower baseline RSA (in boys)
Gordis (2010)	362	52%	12.1y	C	Child abuse	RSA	n/a	Aggression	RPQ	+ for lower baseline RSA (in boys)
Hinnant (2015)	251	49%	8.23y	L	Harsh parenting	RSA	n/a	Delinquent behaviors	PIC2	+ for lower resting RSA (in boys)
Huffman (2020)	101	43%	10.28y	L	Harsh parenting	RSA	n/a	Delinq./agg. behaviors	CBCL	n.s. for basal RSA
McLaughlin (2014)	157	44%	14.9y	C	Child abuse Community violence	RSA	n/a	EB	CBCL	Child abuse: n.s. for basal RSA Community violence: n.s. for basal RSA
Tabachnick (2020)	123	51%	9.46y	L	Child abuse	RSA	n/a	EB	CBCL	+ for higher resting RSA
Basal activity of SNS biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Bubier (2009)	57	51%	7.77y	L	Harsh parenting	PEP	n/a	EB	CSI-4	+ for short baseline PEP
El-Sheikh (2011a)	251	49%	8.23y	L	Marital conflict	SCL	n/a	Delinquency	PIC2	n.s. for baseline SCL
Huffman (2020)	101	43%	10.28y	L	Harsh parenting	PEP	n/a	Delinq./agg. Behaviors	CBCL	Aggression: n.s. for basal PEP Delinquency: + for short basal PEP
McLaughlin (2014)	157	44%	14.9y	C	Child abuse Community violence	PEP	n/a	EB	CBCL	Child abuse: n.s. for basal PEP Comm. viol.: n.s. for basal PEP
Reactivity of PNS biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Bubier (2009)	57	51%	7.77y	L	Harsh parenting	RSA	Socioemotional task and cognitive challenge	EB	CSI-4	n.s. for reactive RSA
Cui (2019)	57	54%	13.19y	C	Parental psych. control Neighborhood violence	RSA	Socioemotional task	Aggressive behavior	PBFS	Par. psych. control: n.s. for reactive RSA Neigh. viol.: + for RSA augmentation
El-Sheikh (2006)	180	49%	9.7y	L	Marital conflict	RSA	Socioemotional task	EB	PIC	n.s. for reactive RSA
El-Sheikh (2011a)	251	49%	8.23y	L	Marital conflict	RSA	Cognitive challenge	Delinquency	PIC2	+ for RSA augmentation (in boys)
El-Sheikh (2011b)	413	47%	8.13y	L	Marital conflict	RSA	Socioemotional task and cognitive challenge	EB	PIC2	n.s. for reactive RSA
Huffman (2020)	101	43%	10.28y	L	Harsh parenting	RSA	Psychosocial evaluative task	Delinq./agg. behaviors	CBCL	+ for RSA withdrawal
Katz (2007)	130	62%	5.04y	C	Marital conflict	RSA	Socioemotional task	Conduct problems	DISC	Father-reported violence: + for RSA augmentation
McLaughlin (2014)	157	44%	14.9y	C	Child abuse Community violence	RSA	Psychosocial evaluative task	EB	CBCL	Child abuse: + for RSA withdrawal Community violence: n.s for RSA reactivity
Obradovic (2011)	260	50%	5.33y	C	Marital conflict	RSA	Socioemotional task and cognitive challenge	EB	HBQ, BPI	+ for RSA augmentation (interpersonal stressor) + for RSA withdrawal (cog. stressor)
Peng (2021)	332	55%	13.7y	L	Marital conflict	RSA	Psychosocial evaluative task	EB	CBCL-YSR	+ for RSA augmentation (speech task) n.s. for reactive RSA (arithmetic task)
Tabachnick (2020)	123	51%	9.46y	L	Child abuse	RSA	Socioemotional task	EB	CBCL	+ for RSA withdrawal
Reactivity of SNS biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Bubier (2009)	57	51%	7.77y	L	Harsh parenting	PEP	Socioemotional task and cognitive challenge	EB	CSI-4	n.s. for reactive PEP
El-Sheikh (2005)	180	49%	9.67y	C	Marital conflict	SCL	Socioemotional task	EB	PIC	+ for high SCL-R (in girls)
El-Sheikh (2007)	157	55%	9.31y	L	Marital conflict	SCL	Socioemotional task and cognitive challenge	EB	PIC	+ for high SCL-R (in girls, argument challenge)
El-Sheikh (2011a)	251	49%	8.23y	L	Marital conflict	SCL	Cognitive challenge	Delinquency	PIC2	n.s. for SCL-R
Erath (2009)	251	49%	8.23y	C	Harsh parenting	SCL	Socioemotional task and cognitive challenge	EB	PIC2	+ for low SCL-R
Erath (2011)	251	49%	8.23y	L	Harsh parenting	SCL	Socioemotional task and cognitive challenge	EB	PIC2	Earlier emergence/persistence of EB: + for high SCL-R (in boys) Later increases: + for low SCL-R (in boys) − for high SCL-R (in girls)
Huffman (2020)	101	43%	10.28y	L	Harsh parenting	PEP	Psychosocial evaluative task	Delinq./agg. Behaviors	CBCL	Aggression: n.s. for reactive PEP Delinquency: n.s. for reactive PEP
McLaughlin (2014)	157	44%	14.9y	C	Child abuse Community violence	PEP	Psychosocial evaluative task	EB	CBCL	Child abuse: n.s. for reactive PEP Comm. viol.: n.s. for reactive PEP
Obradovic (2011)	260	50%	5.33y	C	Marital conflict	PEP	Socioemotional task and cognitive challenge	EB	HBQ, BPI	n.s. for reactive PEP

^aAbbreviations: BPI – Berkeley Puppet Interview; C – cross-sectional; CBCL – Child Behavior Checklist; CBCL-YSR: Child Behavior Checklist-Youth Self-Report; CSI-4 – Child Symptom Inventory; CVI – cardiac vagal index; DISC – Diagnostic Interview Schedule for Children; EB – externalizing behaviors; HBQ – MacArthur Health and Behavior Questionnaire; L – longitudinal; PBFS: Problem Behavior Frequency Scale; PIC: Personality Inventory for Children; PIC2: Personality Inventory for Children-II; RPQ: Reactive-Proactive Questionnaire; R-PRPA – Revised Parent Rating Scale for Reactive and Proactive Aggression; RSA – respiratory sinus arrhythmia; SRASBM: Self-Report of Aggression and Social Behavior Measure

^bMajor findings indicate non-statistically significant (n.s.) estimated moderation effects, and whether estimated associations between the exposure and outcome were positive (+) or negative (−) for sub-groups defined by biomarker measures. Specific exposures and outcomes are specified if more than one was assessed in the study, and any population sub-groups to which significant effects were confined are listed in bold.

Table 4.

Summary of included studies examining neuroendocrine system biomarkers as moderators^a

First author (Year)	n	% male	Mean age	Study design	Childhood violence exposure	Biomarker(s)	Stimuli	Outcome	Outcome measure	Major findingsb
Basal activity of neuroendocrine biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Lipscomb (2018)	361	43%	4.62y	L	Hostile parenting	Salivary cortisol	n/a	EB	CBCL-TRF	n.s. for morning cortisol
Vaillancourt (2018)	198	51%	2.8y	C	Peer victimization	Salivary cortisol	n/a	Phys. agg.	ECHOS, PPVM, CBCL-TRF	+ for lower cortisol levels (in boys) − for higher cortisol levels (in boys)
Reactivity of neuroendocrine biomarker as moderator of relation between childhood violence exposure and externalizing behavior outcome
Cook (2018)	100	32%	15.09y	L	Parental psych. control	Salivary cortisol	Psychosocial evaluative task	EB	YSR	+ for higher cortisol reactivity
Hagan (2014)	88	50%	18.67y	C	Child abuse	Salivary cortisol	Psychosocial evaluative task	EB	ASR	+ for lower cortisol reactivity
Hastings (2011)	402	51%	4.01y	C	Harsh mat. punishment	Salivary cortisol	Socioemotional task	EB	CBCL	+ for higher arrival cortisol (in boys)
Kuhlman (2018)	121	53%	12.77y	C	Child abuse	Salivary cortisol	Psychosocial evaluative and physical discomfort task	EB	CBCL	Emo. abuse: + for higher reactivity Phys. abuse: + for blunted reactivity
Madden (2019)	57	19%	19.47y	C	Child emotional abuse	Salivary cortisol	Psychosocial evaluative task	Dating conflict	CTS2	+ for lower cortisol reactivity
Peckins (2018)	190	100%	12y	L	Marital conflict	Salivary cortisol	Psychosocial evaluative task	AB	SRD	n.s. for cortisol reactivity
Saxbe (2012)	54	54%	15.2y	L	Family aggression	Salivary cortisol	Psychosocial evaluative task	AB	SRA	+ for higher cortisol reactivity

^aAbbreviations: AB – antisocial behavior; ASR – Adult Self-Report; BPAQ – Buss and Perry Aggression Questionnaire; C – cross-sectional; CASI-4R – Child and Adolescent Symptom Inventory-4R; CBCL – Child Behavior Checklist; CBCL-TRF – Child Behavior Checklist-Teacher’s Report Form; CBCL-YSR – Child Behavior Checklist-Youth Self-Report; CTS2 – Conflict Tactics Scale-Revised; DHEA-S – dehydroepiandrosterone-sulphate; DISC-IV – Diagnostic Interview Schedule for Children Revised; EB – externalizing behaviors; ECHOS – Early Childhood Observation System; HCC – hair cortisol concentration; K-SADS – Kiddie Schedule for Affective Disorders and Schizophrenia; L – longitudinal; PEI – Pupil Evaluation Inventory; PPVM – Pre-School Peer Victimization Measure; RPQ – Reactive-Proactive Questionnaire; SDQ – Strengths and Difficulties Questionnaire; SRA – Self-Reported Antisocial Behavior Scale; SRD – Self-Report of Delinquency

^bMajor findings indicate non-statistically significant (n.s.) estimated moderation effects, and whether estimated associations between the exposure and outcome were positive (+) or negative (−) for sub-groups defined by biomarker measures. Specific exposures and outcomes are specified if more than one was assessed in the study, and any population sub-groups to which significant effects were confined are listed in bold.

Table 5.

Summary of included studies examining multi-system biomarkers as moderators^a

First author (Year)	n	% male	Mean age	Study design	Childhood violence exposure	Biomarker(s)	Stimuli	Outcome	Outcome measure	Major findingsb
Basal activity of biomarker(s) as moderator of relation between childhood violence exposure and externalizing behavior outcome
Chen (2015)	425	50%	11.87y	C	Harsh parenting	Salivary cortisol and sAA	n/a	EB	CBCL, YSR	+ for asymm. cortisol/sAA (in boys)
Reactivity of biomarker(s) as moderator of relation between childhood violence exposure and externalizing behavior outcome
Koss (2014)	195	45%	7.99y	L	Marital conflict	Salivary cortisol and sAA	Socioemotional task	EB	CBCL	Concurrent EB: n.s. for cortisol/sAA Longitudinal EB: + for asymmetrical, − for symmetrical cortisol/sAA
McKernan (2018)	153	48%	12.92y	L	Marital conflict	RSA-R and SCL-R	Psychosocial evaluative task	EB	CBCL	+ for co-inhibition − for reciprocal sympathetic activation n.s. for reciprocal parasympathetic activation or co-activation
Philbrook (2018)	252	47%	15.79y	L	Marital conflict	RSA-R and SCL-R	Cognitive challenge	EB	YSR	+ for co-activation
Basal activity and reactivity of biomarker(s) as moderation of relation between childhood violence exposure and externalizing behavior outcome
El-Sheikh (2009) #1	176	44%	8.69y	C	Marital conflict	RSA, RSA-R, SCL, and SCL-R	Socioemotional task and cognitive challenge	EB	PIC2, SBS	Mother-reported delinquency: + for co-inhibition, − for reciprocal activation Father-reported delinquency: + for co-activation
El-Sheikh (2009) #2	251	49%	8.23y	C	Marital conflict	RSA, RSA-R, SCL, and SCL-R	Socioemotional task and cognitive challenge	EB	PIC2	+ for co-inhibition and co-activation
El-Sheikh (2009) #3	150	50%	9.27y	C	Marital conflict	RSA, RSA-R, SCL, and SCL-R	Socioemotional task and cognitive challenge	EB	CBCL, TRF	+ for co-inhibition and co-activation
Gordis (2010)	362	52%	12.1y	C	Child abuse	RSA and SCL-R	Socioemotional task	Aggression	RPQ	+ for co-inhibition and co-activation (in girls)

^aAbbreviations: AB – antisocial behavior; ABF – Antisocial Behaviors Form; C – cross-sectional; CBCL – Child Behavior Checklist; CBCL-TRF – Child Behavior Checklist-Teacher’s Report Form; CBCL-YSR – Child Behavior Checklist-Youth Self-Report; EB – externalizing behaviors; L – longitudinal; PIC2 – Personality Inventory for Children-II; RPQ – Reactive-Proactive Questionnaire; RSA – respiratory sinus arrhythmia; sAA – salivary alpha-amylase; SBS – Student Behavior Survey; SCL – skin conductance level

^bMajor findings indicate non-statistically significant (n.s.) estimated moderation effects, and whether estimated associations between the exposure and outcome were positive (+) or negative (−) for sub-groups defined by biomarker measures. Specific exposures and outcomes are specified if more than one was assessed in the study, and any population sub-groups to which significant effects were confined are listed in bold.

3.2. Studies examining PNS, SNS, and neuroendocrine biomarkers as mediators

As shown in Table 2, one cross-sectional study among adolescents assessing PNS biomarker reactivity to a psychosocial evaluative task failed to find significant mediating effects, with a significant association observed between blunted RSA reactivity and higher levels of externalizing symptoms, but no association between interpersonal violence exposure and RSA reactivity (Busso et al., 2017). Studies investigating the mediating role of SNS biomarker reactivity exhibited mixed results. A cross-sectional study by Timmons et al. (2019) suggested that higher SCL in response to a socioemotional task significantly mediated the link between harsh parenting and dating aggression perpetration among young adult women, but not among men. In contrast, three other studies among children and adolescents did not yield any significant mediating effects of SNS reactivity (using PEP and SCL in response to psychosocial evaluative, cognitively challenging, and socioemotional tasks) in the relation between childhood exposure to violence and externalizing behavior (Busso et al., 2017; El-Sheikh et al., 2007; El-Sheikh, 2005).

Twelve studies examined neuroendocrine biomarkers as potential mediators of the relation between childhood violence exposure and externalizing behavior (Table 2). Eleven of these studies assessed HPA-axis activity using either basal or reactive cortisol levels. Five out of six studies that used basal cortisol collected via saliva or hair yielded non-significant findings for mediation across populations that varied in age from toddlers to adults (Alink et al., 2012; Cima et al., 2008; DePasquale et al., 2019; Doom et al., 2020; Sturge-Apple et al., 2012). However, White and colleagues (2017) found that lower hair cortisol concentration mediated the link between child abuse and higher externalizing behavior. In addition to cortisol, Coelho et al.’s cross-sectional study (2016) tested whether serum copeptin mediated the relation between child abuse and externalizing behavior and yielded null findings. A longitudinal study among toddlers failed to find evidence of basal PNS activity as a mediator of the relation between childhood violence exposure and externalizing behaviors, but did find evidence for mediation by basal SNS activity (Sturge-Apple et al., 2012). Specifically, the authors reported a significant association between maternal harsh parenting and elevated resting CVI, but no association between CVI and externalizing behaviors (Sturge-Apple et al., 2012). They also found that harsh parenting had significant indirect effects on externalizing behavior over time via elevated resting CSI among toddlers who exhibited “hawk-like” phenotypes (i.e., highly active and approachable) but not among those who exhibited “dove-like” phenotypes (i.e., avoidant and inhibited) (Sturge-Apple et al., 2012).

Regarding mediation by biomarker reactivity, three out of five studies that examined HPA-axis reactivity, primarily via salivary cortisol reactivity, found significant mediating effects. In a cross-sectional study among adolescents, Busso et al. (2017) found that blunted HPA-axis (assessed using the ratio of cortisol to DHEA-S) in response to psychosocial evaluative tasks significantly mediated the relation between interpersonal violence and externalizing behaviors. Similarly, Timothy et al.’s (2019) cross-sectional study among boys also found that blunted cortisol in response to psychosocial evaluative tasks mediated the link between interpersonal violence and externalizing behavior. In a longitudinal study, Davies et al. (2007) found that interparental conflict was associated with blunted cortisol in response to a socioemotional task, which in turn predicted increases in externalizing symptoms over a two-year period. However, two cross-sectional studies among adults did not find statistically significant mediating effects of cortisol reactivity to psychosocial evaluative tasks on the link between child abuse and aggression or externalizing behavior (Gowin et al., 2013; Hagan et al., 2014).

Finally, one study examined whether multi-system biomarkers mediated the link between childhood exposure to violence and externalizing behavior (Table 2). In a longitudinal study over 13 years, Shenk and colleagues (2010) found that childhood sexual abuse significantly predicted asymmetrical HPA-axis and PNS activity (measured by salivary cortisol reactivity and RSA activity in response to a cognitive challenge) in late adolescence, which in turn predicted higher levels of antisocial behaviors in young adult women.

3.3. Studies examining PNS and SNS biomarkers as moderators

Thirteen studies evaluated PNS biomarkers as moderators of the relation between childhood violence exposure and externalizing behavior using RSA (either resting, reactive or both) and yielded mixed findings (Table 3). In particular, of the seven studies that examined resting RSA, three studies among children and adolescents found no significant moderating effects of resting RSA on the relation between various childhood violence exposures (harsh parenting, child abuse, and community violence) and externalizing behavior (Bubier et al., 2009; Huffman et al., 2020; McLaughlin et al., 2014). However, three studies found significant moderating effects, such that the relation between childhood violence (including child abuse, marital conflict, and harsh parenting) and delinquent and aggressive behaviors was stronger among those with lower resting RSA, but only among boys (El-Sheikh et al., 2011; Gordis et al., 2010a; Hinnant et al., 2015). Specifically, Gordis et al. (2010a) assessed RSA and SCL as moderators of the effect of maltreatment on aggression in a multisystemic study and found that for boys, only RSA moderated the effect, whereas for girls, both systems interacted (see section 3.5). In contrast, one longitudinal study found that the relation between child abuse risk in infancy and externalizing behaviors at age 9 was stronger among those with higher resting RSA (Tabachnick et al., 2020). Moderation effect sizes were small across all studies.

Evidence for moderation effects of RSA reactivity was similarly mixed. Overall, three longitudinal studies among children found no evidence that RSA in response to social and cognitive stressors moderated the relation between marital conflict or harsh parenting and subsequent externalizing behaviors (Bubier et al., 2009; El-Sheikh and Hinnant, 2011; El-Sheikh and Whitson, 2006). In contrast, two longitudinal studies found a stronger relation between child abuse or harsh parenting and subsequent externalizing behaviors among children and adolescents who exhibited RSA withdrawal in response to cognitively challenging tasks (Huffman et al., 2020; Tabachnick et al., 2020). In the remaining studies, both RSA withdrawal and RSA augmentation exacerbated observed effects, with the nature of the moderation effect dependent on the specific stressor task used to induce RSA reactivity or confined to certain exposures, outcomes, or population sub-groups. For example, in a cross-sectional study of adolescents, neighborhood violence was positively associated with aggressive behavior only among those exhibiting RSA augmentation in response to a socioemotional task; however, these moderating effects were not observed when examining parental psychological control as the exposure of interest (Cui et al., 2019). A cross-sectional study among preschool-aged children found a stronger relation between marital conflict and conduct problems among children who exhibited RSA augmentation in response to a socioemotional task, but only when considering their father’s reported marital violence, not their mother’s (Katz, 2007). Furthermore, El-Sheikh and colleagues (2011) found that the association between marital conflict and externalizing behavior was stronger among children who displayed RSA augmentation after completing a cognitively challenging task, but only among boys.

McLaughlin and colleagues’ (2014) cross-sectional study indicated that childhood abuse was more strongly associated with externalizing behavior among adolescents who exhibited RSA withdrawal in response to psychosocial evaluative tasks; however, RSA reactivity did not moderate associations between community violence and externalizing behavior. Interestingly, in two studies, the moderating effects were dependent on the specific stimuli used when assessing RSA reactivity. In Obradovic et al.’s (2011) cross-sectional study of kindergarten children, the link between marital conflict and externalizing behavior was stronger among children who displayed RSA augmentation in response to an interpersonal stressor but also among children who displayed RSA withdrawal in response to a cognitive stressor. Peng and colleagues (2021) found stronger effects of marital conflict on externalizing behavior across three time points among adolescents who displayed RSA augmentation in response to a speech task, but found no moderating effects of RSA reactivity in response to a mental arithmetic task.

Nine studies examined SNS biomarkers as moderators of the relation between childhood violence exposure and externalizing behavior (Table 3). Among four studies that assessed moderation by resting levels of PEP or SCL, two longitudinal studies found significant moderating effects, such that high levels of harsh parenting significantly predicted subsequent externalizing behavior and delinquency, but not aggression, among children with shorter resting PEP (Bubier et al., 2009; Huffman et al., 2020). However, another longitudinal study found no moderating role for baseline SCL between marital conflict and delinquency over time (El-Sheikh et al., 2011). Similarly, in McLaughlin and colleagues’ cross-sectional study, no moderating role of resting PEP was found in the relations between child abuse and community violence on externalizing behavior (2014).

Among studies assessing the moderating effects of SCL and PEP reactivity, results were somewhat mixed. For studies examining SCL, one longitudinal study found no evidence of moderation by reactive SCL in response to a cognitive challenge task in the relation between marital conflict and delinquency over time (El-Sheikh et al., 2011); however, the remaining studies of SCL reactivity found some evidence for moderation, though inconsistent in direction and across groups defined by sex. Two studies found a positive association between marital conflict and externalizing behavior among girls exhibiting higher SCL in response to cognitively challenging and socioemotional tasks (El-Sheikh et al., 2007; El-Sheikh, 2005). In contrast, a cross-sectional study by Erath and colleagues (2009) indicated that the relation between harsh parenting and externalizing behavior was stronger at lower levels of SCL in response to cognitively challenging and socioemotional tasks. However, the findings were more complicated in a 3-year longitudinal follow-up to this study, in which harsh parenting was associated with earlier emergence and persistence of externalizing behaviors among boys who exhibited higher SCL in response to socioemotional and cognitive stressors, but with later increases in externalizing behaviors among boys who exhibited lower SCL in response to socioemotional and cognitive stressors (Erath et al., 2011). They also found that externalizing behaviors declined over time for girls with greater exposure to harsh parenting who exhibited higher SCL in response to socioemotional and cognitive stressors (Erath et al., 2011).

Four studies examining PEP reactivity found no evidence for reactive PEP moderating the relation between childhood violence exposure and externalizing behavior. Specifically, two cross-sectional studies found no moderating effects of PEP in response to psychosocial evaluative, cognitively challenging, and socioemotional tasks on the association between child abuse, community violence, or marital conflict and externalizing behavior (McLaughlin et al., 2014; Obradović et al., 2011). Similarly, in two longitudinal studies, the authors reported no significant moderating effects of PEP in response to cognitively challenging and psychosocial evaluative tasks (Bubier et al., 2009; Huffman et al., 2020).

3.4. Studies examining neuroendocrine biomarkers as moderators

Nine studies investigated salivary cortisol as a potential moderator of the relation between childhood violence exposure and externalizing behavior (Table 4). Among two studies that assessed basal or diurnal salivary cortisol levels, one found no moderating effects of morning salivary cortisol levels on the relation between harsh parenting and externalizing behaviors (Lipscomb et al., 2018). However, in a cross-sectional study by Vaillancourt and colleagues (2018), peer victimization was positively associated with physical aggression among boys with lower basal cortisol levels, and negatively associated among boys with higher basal cortisol levels. The interactive effects of peer victimization and salivary cortisol were not apparent for girls.

Seven studies measured the moderating role of cortisol reactivity, with mixed results. Only one study found no moderating effects of cortisol reactivity to psychosocial evaluative tasks on the longitudinal relation between marital conflict and antisocial behavior among adolescent males (Peckins et al., 2018). Three studies found stronger associations between childhood violence exposures and externalizing behaviors among those with higher cortisol reactivity, including one cross-sectional study demonstrating that the link between harsh maternal punishment and externalizing behavior was stronger among boys, but not girls, who displayed higher cortisol levels measured upon arrival at the lab (Hastings et al., 2011), and two longitudinal studies in which family violence measures (parental psychological control and cumulative family aggression, respectively) were more strongly associated with externalizing and antisocial behavior among adolescents wither higher cortisol reactivity to psychosocial evaluative tasks (Cook et al., 2018; Saxbe et al., 2012). Another two cross-sectional studies among young adults found stronger associations among those with lower cortisol in response to psychosocial evaluative tasks, including one study of childhood emotional abuse and dating conflict (Madden and Shaffer, 2019) and one of childhood abuse and externalizing behavior (Hagan et al., 2014). Finally, Kuhlman et al. (2018) found disparate results depending on the specific measure of childhood violence exposure. Their cross-sectional study revealed that emotional abuse was associated with more externalizing behavior among children who displayed higher cortisol in response to a physical discomfort task, while physical abuse was associated with more externalizing behavior among children who displayed blunted cortisol in response to a physical discomfort task. Moderation effect sizes were small across all studies.

3.5. Studies examining multi-system biomarkers as moderators

Eight studies investigated the moderating role of multi-system biomarkers on the relation between childhood exposure to violence and externalizing behavior (Table 5). Specifically, two studies assessed the interaction between HPA-axis and SNS activity (represented by salivary cortisol and sAA, respectively) and found some consistent evidence that asymmetrical activity of the HPA-axis and SNS may exacerbate the adverse effects of childhood violence exposures on externalizing behavior. In a cross-sectional study, Chen and colleagues (2015) found that harsh discipline and asymmetrical basal cortisol and sAA significantly predicted externalizing behaviors for boys (but not girls), whereas harsh discipline was not associated with externalizing behavior under symmetrical profiles. Koss and colleagues (2014) did not find any interactive effects between marital conflict, cortisol, and sAA in response to socioemotional tasks on concurrent externalizing behaviors, but did find a positive longitudinal association between marital conflict at age 8 and externalizing behaviors at age 12 among children who displayed asymmetrical cortisol and sAA in response to socioemotional tasks.

Six studies examined interactive effects between childhood exposure to violence, PNS, and SNS activity (using RSA and SCL) on externalizing behavior. Most of these studies provided evidence that co-inhibition and co-activation profiles of PNS and SNS exacerbated the adverse effects of childhood violence exposures. This included two cross-sectional studies by El-Sheikh and colleagues (2009) that assessed resting RSA/SCL and RSA/SCL in response to cognitive challenges and socioemotional tasks, in which marital conflict was associated with higher levels of parent- and teacher-reported delinquency among children who exhibited co-inhibition (e.g., low RSA and low SCL-R) and co-activation (e.g., high RSA and high SCL-R), but was not associated among children who exhibited reciprocal activation (e.g., high RSA and low SCL-R). In another cross-sectional study, Gordis and colleagues (2010) similarly found that childhood maltreatment was significantly associated with aggression among girls who displayed co-activation and co-inhibition patterns of resting RSA and SCL in response to a socioemotional task, but was not associated among girls who displayed reciprocal patterns or among boys. In a longitudinal study among adolescents, Philbrook and colleagues (2018) found that marital conflict predicted increased externalizing behaviors over time among adolescents with co-activation of PNS and SNS, but was not associated among adolescents who displayed reciprocal or co-inhibition patterns. In contrast, McKernan and Lucas-Thompson (2018) found that marital conflict was positively associated with increased externalizing behavior over time among those who exhibited co-inhibition of PNS and SNS, but was negatively associated among those who exhibited reciprocal sympathetic activation (e.g., low RSA reactivity and high SCL reactivity to a psychosocial evaluative task), and was not associated under conditions of co-activation or reciprocal parasympathetic activation (e.g., high RSA reactivity and low SCL reactivity). Finally, in a third cross-sectional study by El-Sheikh and colleagues (2009), the pattern of associations between marital conflict and delinquency varied by both PNS and SNS activity profiles and by outcome reporter. Specifically, marital conflict was positively associated with mother-reported delinquency among children who exhibited co-inhibition, and was also positively associated with father-reported delinquency, but only among children who exhibited co-activation. Reciprocal patterns of PNS and SNS activity (e.g., higher resting RSA/RSA reactivity and lower SCL reactivity, or lower resting RSA/RSA reactivity and higher SCL reactivity) were protective against parent-reported externalizing behavior after exposure to marital conflict.

4. Discussion

In this systematic review of 46 independent studies of the role of physiological biomarkers in the relation between childhood violence exposure and externalizing behavior, we found some consistent evidence for blunted HPA-axis reactivity as a mediator in this relation, and for PNS/SNS co-inhibition or co-activation as moderators exacerbating this relation. Both higher and lower reactivity of PNS, SNS, and neuroendocrine biomarkers in responses to stressors increased the risk of externalizing behaviors or protected against the development of externalizing behaviors after exposure to childhood violence, consistent with some theories that emphasize that both high and low stress reactivity can be adaptive depending on one’s early environment.

4.1. Summary of results for physiological biomarkers as mediators

We found limited consistent evidence that blunted HPA-axis reactivity may mediate the relation between childhood violence exposure and externalizing behaviors. This is consistent with previous work that has identified decreased HPA-axis reactivity as characteristic of adults with early adverse exposures (Bunea et al., 2017) and that has linked blunted HPA-axis reactivity to adverse behavioral outcomes (Turner et al., 2020). In contrast, studies that examined diurnal or basal cortisol as mediators consistently yielded null results. This may imply that childhood violence exposure is more likely to contribute to dysregulation in HPA-axis responsivity within the context of stress, which in turn leads to more externalizing behavior (Hartman et al., 2013). However, a stronger association between cortisol reactivity (vs. basal cortisol) and externalizing behaviors has not been consistently found across studies (Alink et al., 2008; Figueiredo et al., 2020). Moreover, only one study assessed multi-system biomarkers as potential mediators (Shenk et al., 2010), with significant mediation by asymmetrical HPA-axis and PNS reactivity in adolescence.

Few studies have assessed the potential mediating role of PNS and SNS biomarkers in the relation between childhood violence exposure and subsequent externalizing behavior. Of those, most had null findings, as previously documented in a review of ANS responsivity as a mediator of the association between childhood maltreatment and psychopathology, including externalizing symptoms (Young-Southward et al., 2020). Several reasons may explain the dearth of mediation studies in this area. While the allostatic load theory serves as a guiding framework to design mediation studies, allostatic load is a broad measure that may reflect repeated stressful exposure or a lack of adaptation or inadequate stress response (McEwen & Stellar, 1993). The broad construct of allostatic load and its complex impact on physiological stress systems make measuring psychophysiological dysregulation and changes in stress response systems over time very challenging. Furthermore, a developmental approach is needed to investigate the causal pathways underlying mediation models. Ideally, biomarkers should be measured longitudinally, including multiple time points prior to and after exposure to childhood violence. In addition, since allostatic load represents the interplay of several systems (McEwen & Stellar, 1993), this construct should be measured by a composite score derived from multiple biomarkers across different stress response systems (Seplaki et al., 2005). This intricate multisystem mediation research design can be very costly and hard to execute. Nevertheless, well-designed mediation studies are needed to expand the literature and inform intervention and prevention planning.

4.2. Summary of results for physiological biomarkers as moderators

The link between childhood violence exposure and externalizing behavior was consistently stronger at lower values of resting RSA and RSA reactivity. Notably, lower resting RSA levels, which may represent a restricted ability to emotionally regulate and respond to adverse environmental demands (Bornstein and Suess, 2000; Fortunato et al., 2013), may particularly increase risk among boys (El-Sheikh et al., 2011; Gordis et al., 2010a; Hinnant et al., 2015). The effects of physiological regulation may be more apparent for boys since girls are found to have better cognitive and emotional self-regulation compared to boys (Brody, 2000). RSA augmentation reflects increased PNS activity in response to stress and has been conceptualized to increase insensitivity to punishment (Scarpa and Raine, 2004). Limited consistent findings also revealed the moderating role of RSA withdrawal. Typically, vagal withdrawal promotes increased sympathetic activation in response to stress. However, extreme levels of vagal withdrawal may suggest reduced self-regulation abilities (Hastings and Miller, 2014; Kahle et al., 2018). The combination of exposure to violence, low emotion regulation in response to the environment from lower resting RSA, insensitivity to punishment from RSA augmentation, and low self-regulation from high levels of RSA withdrawal may increase risk for externalizing behavior.

Furthermore, results indicated that both high and low cortisol reactivity may exacerbate the association between childhood violence exposure and externalizing behaviors. Individuals who display high cortisol reactivity may be sensitized to violence exposure and struggle to regulate their conflict-related arousal, increasing their risk of displaying externalizing behaviors (Raine, 2005; van Goozen et al., 2000). Moreover, the HPA-axis may adapt to prolonged hypersecretion of cortisol (i.e., greater cortisol reactivity) under conditions of constant violence exposure, resulting in hyposecretion of cortisol (i.e., lower cortisol reactivity) (Trickett et al., 2010). Thus, individuals with low cortisol reactivity may be under-aroused and more likely to seek out stimulation and arousal through externalizing behaviors when exposed to violence (Raine, 2005; van Goozen et al., 2000). In addition, studies utilizing a multi-system approach reliably found that non-reciprocal coordination of PNS and SNS (i.e., co-inhibition or co-activation) confers risk for externalizing behavior among participants who were exposed to childhood violence. Consistent with Berntson et al.’s doctrine of autonomic space (Berntson et al., 1991), non-reciprocal coordination may confer risk for poor childhood outcomes in adverse environments as these systems dampen or cancel each other’s effects.

4.3. Differential results by study design, stimulus task, and other factors

Overall, a higher proportion of mediation results reported in longitudinal studies were identified as statistically significant, compared to cross-sectional studies. Longitudinal studies are better suited to mediation analysis given their ability to establish the temporal ordering of measures. However, evidence for significant moderating effects of physiological biomarkers was similar from both cross-sectional and longitudinal studies. With respect to different age groups, both mediation and moderation results were more consistently statistically significant in studies that assessed outcomes in adulthood, though these made up a smaller proportion of the included studies. This is consistent with long-term effects of childhood violence exposures on outcomes in adulthood and deserves further exploration. Physiological biomarkers were consistently found to be significant moderators/mediators in the link between violence exposure in the family setting (e.g., marital conflict, child abuse, harsh parenting) and externalizing behaviors, though the observed effects of violence in non-family settings (e.g., exposure to community violence) were also moderated by physiological functioning. Furthermore, measures of externalizing symptoms may encompass a variety of behaviors. Studies focusing on narrower outcome measures (e.g., aggression) more consistently identified significant moderating effects of physiological biomarkers than those focused on the broader construct of externalizing behaviors as a whole.

Finally, the heterogeneity of tasks used to elicit physiological reactivity may contribute to the differential findings. Specifically, acute stress tasks such as cognitive challenges (e.g., mirror tracing), psychosocial evaluative tasks (e.g., Trier Social Stress Test), physical discomfort (e.g., cold pressor test), and socioemotional stressors (e.g., watching or listening to emotional content) were used in the included studies. All tasks were commonly used among children, except for psychosocial evaluative tasks, which were mostly implemented among older children (11-12 years), adolescents, and adults. Notably, several studies conducted among children utilized a combination of cognitive challenge and socioemotional tasks to evoke reactive responses. The pattern of stress task choices for different developmental stages suggests that children (regardless of level of functioning and trait) needed age-appropriate stressors. While the diversity in task choices can help to establish some degree of convergent validity, the variety of tasks lead to different responses and findings. Nevertheless, the lack of “rule of thumb” in choosing stress tasks is an apparent issue in the psychophysiological literature (National Advisory Mental Health Council Workgroup on Tasks and Measures for Research Domain Criteria (RDoC), 2016). As selecting stress tasks that are relevant for samples with certain age, trait, and functioning is crucial, the field needs to establish normative data to understand the basic psychometric properties of popular stress tasks.

4.4. Support for biological theories linking childhood violence exposure to externalizing behaviors

Our results provided support for both the BSC (Boyce and Ellis, 2005) and ACM (Del Giudice et al., 2011) theories. Specifically, the findings on low resting RSA, high SCL reactivity, and high cortisol reactivity influencing the relation between childhood violence exposure and externalizing behaviors are consistent with the BSC theory, such that individuals with high stress responsivity may be at risk for the adverse effects of negative environments (e.g., high childhood violence exposure) but benefit from positive environments (e.g., low childhood violence exposure). Notably, most results aligning with the BSC are from children, suggesting that high stress responsivity during childhood may contribute to differential susceptibility to externalizing behaviors depending on the environment.

ACM posits that changes in the stress response system (hypo- or hyper-responsivity) resulting from adaptation to early life environments may have long-term consequences for developmental outcomes. Therefore, studies finding both RSA augmentation and withdrawal, high and low SCL reactivity, and high and low cortisol reactivity among children influencing such relations lend support to the ACM. However, we are not able to determine if observed physiological functioning reflects adaptations to the early life environment given the lack of examination of mediating processes. ACM also specifically highlights sex differences in stress response system functioning. Several moderation studies found that both sex of children and physiological biomarkers influence the association between childhood violence exposure and externalizing behavior. Future studies should further delineate the interplay of sex differences, physiological responses, and childhood violence exposure on externalizing behavior among adolescents and adults.

Overall, we found that the BSC and ACM theories were mainly supported among children when using specific biomarkers such as RSA, SCL, and cortisol. The current findings help to extend Liu’s (2004) biosocial model of childhood externalizing behavior by considering both mediating and moderating processes that increase the risk of externalizing behavior throughout different developmental phases (childhood through adulthood).

4.5. Limitations

This review has some limitations. Although we focused on externalizing behavior, the included studies varied in the types of externalizing behavior assessed and how they were measured. Heterogeneity also emerged across the included studies in the specific childhood violence exposures, the age range of participants, and the measurement model tested. Such heterogeneity precluded our ability to summarize trends and compare across studies. Additionally, we opted to exclude studies using broader measures of childhood adversity in order to retain the focus on childhood violence exposures; however, violence exposures commonly co-occur with other adverse childhood experiences and research in this area increasingly utilizes composite adversity measures (Appleton et al., 2017). Therefore, we may have missed recent relevant evidence on the mediating and moderating role of physiological biomarkers in the relation between childhood adversity, including violence exposures, and subsequent externalizing behaviors.

In addition, the included studies that assessed physiological reactivity utilized a variety of stress tasks (e.g., social evaluative stressors, cognitive challenge, interpersonal arguments) to elicit reactivity responses. These sources of heterogeneity hindered our effort to synthesize findings across studies and precluded use of a meta-analytic approach to summarize the results. Nevertheless, our review highlighted the dependence of study results on certain aspects of the study design, with important implications for future research, outlined below. Furthermore, this review included more studies that were conducted among children and adolescents (n=40) than adults (n=6), limiting our ability to draw inferences about the mediating or moderating role of physiological biomarkers on the relation between childhood violence exposure and externalizing behavior across different developmental periods. Finally, the focus of this review was only on traditional physiological measures from the ANS and HPA-axis. Other biomarkers related to gene expression and epigenetics could also be important to consider as mediators or moderators in the relation between childhood violence exposure and externalizing behavior (Byrd and Manuck, 2014; Weeland et al., 2015).

5. Conclusions

Despite these limitations, our review suggests that physiological functioning, especially resting RSA, RSA reactivity, cortisol reactivity, and non-reciprocal PNS and SNS activation, moderates the relation between childhood violence exposure and externalizing behaviors. We also identified limited evidence that physiological biomarkers, especially cortisol reactivity, are involved in the causal pathway between childhood violence exposure and externalizing behavior.

5.1. Clinical implications of results

These results have clinical implications for study design, intervention development, and screening. First, physiological biomarkers should be included as objective parameters in randomized clinical trials to better understand treatment or intervention effects. Indeed, growing evidence has indicated that physiological biomarkers are predictors or moderators of the effects of psychotherapy and other interventions for externalizing disorders (Bagner et al., 2012; Beauchaine et al., 2013). Second, specific targeted interventions should be developed to treat individuals with different levels of autonomic functioning and environmental contexts. For example, interventions that aim to focus on improving one’s ability to regulate physiological arousal may be effective ways to buffer the adverse effects of childhood violence exposure (Slopen et al., 2014). Third, screening children in primary care and other settings with respect to their physiological functioning and their exposure to violence may capture those at highest risk for externalizing behaviors, creating opportunities to provide them with more effective interventions (e.g., emotion regulation strategies or parenting training). For example, screening for adverse childhood experiences (ACE) (Finkelhor, 2018) may be supplemented with assessments of physiological functioning.

5.2. Future research directions

Nevertheless, more research is needed to assess the specific conditions under which different physiological biomarkers act as mediators and/or moderators of the relation between childhood violence exposure and externalizing behaviors. Our review revealed differential findings among studies that presented stratified results by demographic variables (Chen et al., 2018; Erath et al., 2011; Gordis et al., 2010a; Murray-Close et al., 2014; Vaillancourt et al., 2018), type of childhood violence exposures (Cui et al., 2019; Kuhlman et al., 2018; McLaughlin et al., 2014), and type of stressors used to induce physiological reactivity (Obradović et al., 2011; Peng et al., 2021). Rather than including demographic variables only as control variables, future research should explicitly compare results according to demographic characteristics and across different types of exposures. Since individuals typically face many types of stressors out of the laboratory context (Obradović et al., 2010), future studies should also assess physiological reactivity to multiple stressors and conduct separate analyses using these indices. Furthermore, studies should include both resting and reactive states of biomarkers, as they are complementary but distinct indices.

As noted above, more research is particularly needed to assess physiological biomarkers as mediators in the relation between childhood exposure to violence and externalizing behavior. Moreover, future studies should test both moderating and mediating effects of physiological biomarkers simultaneously by using moderated mediation and multilevel modeling techniques (e.g., Fagan et al., 2017) to elucidate a more comprehensive understanding of such relations. In addition, longitudinal studies from adolescence to adulthood should be conducted to provide a more comprehensive picture of when and how physiological biomarkers may cause or influence the effects of childhood violence exposure on subsequent behavioral health outcomes.

Few included studies employed a multi-system assessment (PNS and SNS or SNS and HPA-axis) of physiological biomarkers in relation to childhood violence and externalizing behavior. Future research should simultaneously examine multiple systems, including more than two systems (e.g., PNS, SNS, and HPA-axis) as this helps to provide a more comprehensive understanding of how these systems work together to influence developmental outcomes (Bauer et al., 2002). Furthermore, since different biomarkers may be subject to different systematic and random errors, future work should include multiple markers of the same system (e.g., assessing SCL, sAA, and PEP as markers of SNS concurrently) when examining the relation between childhood violence exposure and externalizing behaviors to verify the functional equivalence and appropriate contexts for use of these indices. However, different biomarkers have different properties and researchers need to choose markers appropriate to the research question. Even different biomarkers of the same system have different applications. For example, cortisol extracted from hair, serum and saliva all provide different information over different sampling periods and may be suited to different questions. Saliva cortisol is better for examining acute responses, whereas urinary cortisol allows for integration over a longer period of time, such as 24 hours (El-Farhan et al., 2017).

While both environmental and biological factors can contribute to developmental outcomes, the manner in which these factors interact may greatly vary and thus call for continued investigation. Importantly, the presence of a significant interaction in moderation analyses is not sufficient to determine the manner of sensitivity to environmental influences (Dearing and Hamilton, 2006). Moderation analyses should further probe the effects of individual differences (marked by physiological responses) on sensitivity to environmental influences using region of significance testing (Roisman et al., 2012). Nonetheless, future studies should form specific theory-guided hypotheses before probing such effects to provide insight into optimal strategies in preventing long-term adverse consequences of childhood violence exposures.

Highlights.

• Childhood violence exposure is an established predictor of externalizing behavior.

• Physiological biomarkers may moderate or mediate this relation.

• We found more evidence of physiological biomarkers as moderators than mediators.

• High and low reactivity, and non-reciprocal SNS/PNS activation, increased risk.

• Findings varied by type of biomarker, assessment methods, and participant sex.