Links between Oppositional Defiant Disorder Dimensions, Psychophysiology, and Interpersonal versus Non-interpersonal Trauma

Abstract

The etiology of oppositional defiant disorder (ODD) is not well understood but appears to have both biologically-based roots and can develop following adverse experiences. The current study is the first to examine the interaction between biologically-based factors and type of trauma experience (i.e., interpersonal and non-interpersonal) and associations with ODD. The psychophysiological factors included baseline resting heart rate, respiratory sinus arrhythmia (RSA), and cortisol. ODD was measured as two dimensions of irritable and defiant/vindictive. The sample included 330 children, 3-7 years-old, oversampled for a history of trauma. Results showed the interactions between baseline physiological arousal variables and trauma type in predicting ODD dimensions were not supported. However, the baseline RSA by trauma interaction was a significant predictor of defiance/vindictiveness among boys, but not girls, when interpersonal trauma was compared to controls. Several other gender differences emerged. Among boys, both interpersonal and non-interpersonal trauma were predictive of ODD dimensions; however, among girls, non-interpersonal trauma was not. Among girls, there was a significant negative bivariate relationship between baseline cortisol and irritability. Also, when the sample was restricted to those with interpersonal trauma only and controls, baseline RSA was negatively associated with irritability in girls only (controlling for trauma). Finally, retrospective reports revealed that children who met criteria for ODD diagnosis and experienced interpersonal trauma were more likely to exhibit ODD symptoms prior to their trauma compared to those who experienced non-interpersonal trauma. Results are discussed in the context of previous mixed findings, and avenues for future research are highlighted.

There is a well-documented relationship between childhood externalizing problems (including oppositional defiant disorder [ODD]) and exposure to traumatic events, such as domestic violence (S. E. Evans et al., 2008) and maltreatment (Milot et al., 2010), and other types of non-traumatic adverse events, such as harsh parenting (Gershoff, 2002). This is also evidenced in the high comorbidity between posttraumatic stress disorder (PTSD) and ODD, even among young children (Løkkegaard et al., 2017; Scheeringa & Zeanah, 2008). Further, it appears that interpersonal trauma, such as physical abuse, sexual abuse, and domestic violence are more strongly associated with externalizing problems in children than non-interpersonal trauma, like severe accidents, injuries, or illness (Ford et al., 2011; Kisiel et al., 2009). Generally, children who experience interpersonal trauma also tend to possess more problems with anger, emotional regulation, and disruptive behavior, and fewer strengths than children who experience non-interpersonal trauma (Kisiel et al., 2009). Further, it has been theorized that children with psychopathology who, for example, present a greater parenting challenge, may be at greater risk for victimization (Cuevas et al., 2010). In terms of ODD specifically, it has been shown that among 6-17-year-old children, ODD was associated with victimization, but not with accident or illness related trauma (Ford et al., 1999). In a large community sample of 4-18-year-olds, Price et al. (2013) showed that direct trauma (experiencing trauma personally, as opposed to witnessing) and interpersonal trauma involving immediate family were significantly associated with ODD, but interpersonal trauma involving individuals outside the immediate family and non-interpersonal trauma were not.

While there is a relationship between trauma and ODD, the factors that determine which children who experience trauma either have or go on to develop ODD or other externalizing behaviors are unknown. It is possible that pre-existing, biologically based individual differences put some children at risk for developing psychopathology in the context of trauma, and act as a buffer against the development of psychopathology for others (Cipriano et al., 2011). Several biologically based variables involved in stress responding have been studied in relation to ODD, including those indexing the functioning of the autonomic nervous system (ANS) and the hypothalamic-pituitary-adrenal (HPA) axis, though the findings have been mixed.

The ANS is made up of two branches: the sympathetic nervous system (SNS), and the parasympathetic nervous system (PNS). In response to a stressor, activation of the SNS prepares the body for “fight or flight” by increasing heart rate, dilating blood vessels to skeletal muscle, and activating sweat glands, for example. The PNS is a separate inhibitory system which serves as a “vagal brake” by decreasing heart rate and promoting homeostasis once a stressor recedes, or when the environment is deemed safe (Porges, 2007).

Heart rate (HR) is an index of physiological arousal that is influenced by both the SNS and PNS. Resting heart rate is thought to reflect baseline levels of arousal, with low resting HR indicating hypo-arousal. Many studies have found that lower HR is related to conduct problems(e.g., Fanti et al., 2019; Lorber, 2004; Ortiz & Raine, 2004), but few studies have examined the relationship between resting HR and ODD. Studies that included a combined ODD/CD group(de Wied et al., 2009, Posthumus et al., 2009; Snoek et al., 2004; van Goozen et al., 1998) show mixed results. Schoorl et al. (2016a) found higher resting HR in 66 8-12 year old boys with an ODD diagnosis (and some comorbidities) compared to a non-clinical control group (n = 36).

Some have suggested that one reason for mixed findings for resting HR is because HR is influenced by both the PNS and SNS. Resting respiratory sinus arrhythmia (RSA), on the other hand, is a measurement of variability in beat-to-beat intervals, and is an index of the PNS, or vagal influence on the heart (Porges, 1995). Baseline RSA is thought to reflect a capacity for self-regulation, with low baseline levels indicative of dysregulation (Beauchaine & Thayer, 2015). Several studies by Beauchaine and colleagues have shown that children high on ODD/CD or low on prosocial behavior had lower baseline RSA (Beauchaine, 2001; Beauchaine et al., 2007; Beauchaine et al., 2013). Although the relationship between baseline RSA and ODD specifically has been studied less often, some have failed to find an association (Crowell et al., 2006; Vidal-Ribas et al., 2017). However, in a longitudinal study of a community sample, Wagner et al. (2017) found that lower baseline RSA measured across infancy (3-24 months of age) was related to higher levels of ODD in first grade. Thus, similar to the findings on HR, the findings on the relationship between baseline RSA and ODD have been inconsistent.

The HPA axis is another component of the stress response system in addition to the ANS. It can be thought of as a sequence of hormone releases that results in the production and elevation of cortisol. The effects of this cortisol elevation serve to prepare the body for action, similar to the SNS, but it also has longer-term regulatory influence (Del Giudice et al., 2011). A meta-analysis conducted by Alink and colleagues (2008) examined the relationship between baseline cortisol and externalizing problems across 72 study outcomes. Results showed a small but significant effect, with lower baseline cortisol associated with higher externalizing behaviors. However, when age was tested as a moderator, preschoolers (age 0-5) showed the opposite effect, with higher baseline cortisol predicting higher externalizing behaviors. In terms of ODD/CD and ODD specifically, Schoorl et al. (2016b) found lower baseline cortisol levels in boys with ODD/CD without anxiety compared to those with ODD/CD/anxiety, and healthy controls. In contrast, Snoek et al. (2004) found no differences between children with ODD/CD and normal controls in a mixed-sex sample of 7-12 year-old children. van Goozen et al. (1998) found lower baseline cortisol levels in 8-11-year-old boys with ODD compared to normal controls.

The mixed findings regarding the relationship between baseline ANS and HPA functioning and externalizing behaviors may be, in part, due to gender differences. For example, Beauchaine et al. (2008) showed that boys (ages 8-12) who scored high on aggression showed lower baseline RSA than those who scored low on aggression, but there was no significant difference among girls. Dietrich et al. (2007) found that preschool externalizing behaviors were related to higher baseline RSA at ages 10-13, but only among girls. In terms of cortisol, a study of 6-11 year-old children found a significant gender by group interaction such that boys with disruptive behavior disorders and girls without (healthy controls) had the lowest cortisol levels (Dorn et al., 2009). Post-hoc analyses showed that boys with ODD had lower cortisol than girls with ODD. In a longitudinal study of 6-16 year-old children followed for 3 years, Shirtcliff et al. (2005) also found that boys with more externalizing problems had lower trait cortisol, but this difference was not significant among girls.

Relationship with Trauma

It has been suggested that physiological arousal may act as a risk- or protective-factor for the development of externalizing behaviors in the context of adversity (Eisenberg et al., 2012; El-Sheikh et al., 2009). Two studies showed that an interaction effect between baseline RSA and trauma exposure was associated with internalizing behaviors, but not externalizing (Gray et al., 2017; McLaughlin et al., 2015). There are no other known studies that have examined all three constructs of true trauma exposure, baseline physiological arousal, and ODD in children simultaneously.

Understanding whether baseline level of physiological arousal, in conjunction with trauma exposure, acts as a risk or protective factor for the development of ODD is contingent on knowing the temporal relations between physiological arousal level, trauma exposure, and the onset of ODD symptoms. The two prior studies on RSA, trauma, and externalizing (Gray et al., 2017; McLaughlin et al., 2015) were cross-sectional studies. While cross-sectional studies provide important preliminary evidence about the strength of relations between variables, they cannot inform causal relations. The limited longitudinal studies which have been conducted have assessed environmental quality or harsh parenting rather than trauma (Eisenberg et al., 2012; Hinnant et al., 2015). Understanding causal relations is also important because of the increasing evidence that trauma does not happen at random, and those who experience trauma may have pre-existing vulnerabilities (Danese, 2020). For example, a recent review demonstrated that cognitive deficits that had often been causally attributed to maltreatment actually tend to pre-exist experiences of maltreatment (Danese et al., 2017). If a proportion of ODD shows a similar relationship with trauma exposure, this would have implications for causal theories about risk and protective factors.

ODD symptom dimensions

In the research examining relationships between ODD and trauma exposure or psychophysiological indices described above, ODD was conceptualized as a unitary construct. However, several studies have supported the multidimensionality of ODD. Although several two- and three-factor models have been proposed (Aebi et al., 2010; Burke et al. 2010; Burke et al., 2014; Rowe et al., 2010; Stringaris & Goodman, 2009), a large study comparing models concluded that a bifactor model with two correlated specific factors fit the data best (Burke et al., 2014). In this model, the irritable factor was comprised of three symptoms (often losing temper, touchy, angry), and the defiant/vindictive factor included the remaining five symptoms (argues, defies, annoys, blames, spiteful/vindictive). In addition to the factor analytic support, the dimensions of ODD have also been shown to have differing correlates and longitudinal outcomes (Mikolajewski et al., 2017; Ollendick et al., 2018; Rowe et al., 2010). Thus, it is possible that the mixed findings on the relationships between HR, RSA, and cortisol and ODD are inconclusive because the multidimensionality of ODD was not accounted for.

Current Study

Previously conducted research consistently shows that experiencing trauma is related to symptoms of ODD (Carliner et al., 2017; Ford et al., 1999; Scheeringa, 2015; Scheeringa et al., 2011), and some work suggests this relationship is specific to interpersonal trauma (Ford et al., 1999). Scheeringa (2015) found repeated trauma was associated with ODD using the same sample as this study but did not specifically examine trauma type (e.g., interpersonal versus non-interpersonal).

The current study aimed to examine the interaction between baseline arousal variables and the type of trauma experience (interpersonal versus non-interpersonal) in the prediction of ODD symptomatology. In line with research which has supported the multidimensionality of ODD symptoms, irritable and defiant/vindictive dimensions of ODD were examined. It was predicted that in the context of interpersonal trauma, higher ODD symptoms (particularly defiance/vindictive symptoms), would be associated with lower resting levels of HR, lower resting RSA, and low baseline cortisol. Given evidence of possible gender differences between physiological arousal and externalizing, males and females were examined separately. While the primary research aim was to compare interpersonal to non-interpersonal trauma, the availability of an age-matched non-trauma-exposed control group allowed additional analyses to provide a more complete context of how trauma-exposure associates with physiological variables and ODD. In addition, given the importance of determining the temporal relations of these variables and whether ODD symptoms represent pre-existing conditions or consequences of trauma exposure, timing of ODD onset in relation to trauma exposure will be examined.

Method

Participants

The current study is a secondary analysis of previously collected data (R01MH065884). Children who participated in the original study on posttraumatic stress disorder (PTSD) in early childhood were recruited based on their trauma history. Thus, children were recruited through a Level 1 Trauma Center registry, newspaper advertisements, and women’s shelters (N = 284) in the New Orleans metro area. To be eligible, children had to be between 36 and 83 months old at the time of enrollment, and they had to have experienced at least one life-threatening trauma that occurred when the child was at least 3 years of age. Children were excluded if they experienced a significant head trauma (Glascow Coma Scale score seven or less), were diagnosed with mental retardation or autistic disorder, suffered from blindness or deafness, or were from non-English speaking families. Additionally, a control group (N = 46) was recruited by asking participants for neighbors and acquaintances with children within the specified age range. Inclusion criteria was the same for the control group, except they had no experience of trauma. The 330 children who participated in the study were between the ages of 3 and 7 years-old (M = 5.06 years), and 61.8% (N = 204) were male. Caregivers reported the race/ethnicity of the children to be 67.6% Black, 18.8% White, 7.0% mixed, 2.1% other, and 4.5% Latino. Sixty-six families did not return for the second session, leaving 264 who attempted to participate in the physiological data collection (223 trauma-exposed; 41 controls). Heart rate and RSA data were not collected on 14 children due to technical problems or refusal to wear electrodes, leaving 250 children with data (211 trauma-exposed, 39 controls). Saliva was not collected on 14 children who refused to cooperate, and 6 samples were unable to be assayed, leaving 244 children with cortisol data (204 trauma-exposed 40 controls). For one child, it could not be determined if the trauma they experienced qualified as interpersonal or non-interpersonal, and two children were missing data on ODD variables.

Procedure

The study was approved by the Tulane University Institutional Review Board. Data were collected over two laboratory sessions. Only the caregiver attended the first session lasting approximately 2.5 hours. Written informed consent was obtained from the caregiver upon their arrival. They then participated in a psychiatric interview about their child and completed several questionnaires. The child accompanied their caregiver to the second session, which occurred approximately one week later and lasted approximately 1.5 hours. During the second session, interviews were conducted with the child, and physiological data were collected. Families were compensated for their participation following each session.

Measures

Structured Diagnostic Interview.

Caregivers were interviewed about their child’s symptoms across nine psychiatric disorders using the Preschool Age Psychiatric Assessment (PAPA; Egger et al., 2006; Egger & Angold, 2004). However, only data from the ODD and PTSD modules were used in the current study.

The ODD module of the PAPA is used to assess DSM-IV symptoms using developmentally appropriate prompts that pertain to preschool aged children. Each ODD symptom was coded as either absent or present. Two subscales of ODD were created because research has suggested that an irritable dimension differs from a defiant/vindictive dimension on correlates potentially related to baseline physiological arousal such as aggression and emotionality, as well as risk for later psychopathology (S. C. Evans et al., 2016; Ezpeleta et al., 2012; Mikolajewski et al., 2017). The three symptoms of angry/irritable mood (loses temper, touchy/easily annoyed, angry/resentful) were summed to create the “irritable” outcome variable in the current study. The four symptoms of argumentative/defiant behavior (argues with adults, defies, annoys others, blames others) and the one symptom of vindictiveness were summed to create the “defiant/vindictive” outcome variable.

The assessment of trauma from the PTSD module of the PAPA was used to determine what types of traumas children experienced. Interpersonal trauma was defined as (1) witnessing another person being beaten, raped, threatened with serious harm, shot at, seriously wounded, or killed, (2) physical abuse, (3) sexual abuse, sexual assault, or rape, or (4) being kidnapped. Non-interpersonal trauma was defined as (1) being in an accident/crash with an automobile, plane, or boat (2) being attacked by an animal, (3) man-made disasters such as fire or war, (4) natural disasters, (5) accidental burning, (6) near drowning, or (7) hospitalization, emergency room visit, and/or invasive medical procedures. Three trauma related variables were created. A binary trauma type variable was created to indicate whether a child experienced interpersonal trauma (coded 1) or non-interpersonal trauma (coded 0; controls were excluded). Additionally, an interpersonal trauma variable was created to indicate whether a child experienced interpersonal trauma (interpersonal coded 1; controls coded 0). Similarly, a non-interpersonal trauma variable was created to indicate whether a child experienced non-interpersonal trauma (interpersonal coded 1, controls coded 0). Children who reported both interpersonal and non-interpersonal trauma were excluded from analyses. Including these children could obscure differences between groups. Further, this group of children may be distinct (e.g., experiencing more trauma than other children), but was not a large enough group in the current sample to perform separate analyses.

Psychophysiological Assessment.

The psychophysiological assessment occurred during the second session, which was scheduled to begin between 9:00-9:30am. After a 20-30 minute period of acclimation to the lab, the child provided a saliva sample from which baseline cortisol level (nmol/l) was determined. A second cortisol sample was collected later in the assessment but was not included in the current study. Saliva was collected by asking the child to chew gently on a cotton roll. After one minute, the cotton roll was removed, placed in a labeled Salivette tube, and stored in a freezer. The samples were later shipped to and assayed by the laboratory of Dr. Clemens Kirschbaum in Dresden, Germany.

Following the saliva collection, the electrocardiogram data were collected and analyzed using equipment and software from the James Long Company. Data were collected using three electrodes which were placed on the child (left rib cage, right rib cage, and abdomen). After electrodes were placed, children were asked to sit still and watch a 45 s neutral video of old-fashioned biplanes, with relaxing music in the background. Using a neutral video stimulus to obtain baseline data allows for standardization across participants. Additional videos were also presented, but data collected during those videos were not included in the current analyses.

Data were passed through a bioamplifier with bandpass filtering set at 0.1 and 1000 Hz. A multiple-pass, self-scaling algorithm was used to automatically detect R waves; however, R waves were also visually inspected during scoring and corrected if necessary. The current analyses examined mean interbeat intervals (IBI), which were calculated as the mean distance between successive R waves during the 45 s neutral video. Spectral analysis using discrete Fourier transformation was then used to derive respiratory sinus arrhythmia (RSA) from prorated and detrended IBIs. Age-appropriate frequency bands were used based on previous research (3 year-olds: 0.30–0.75 Hz; age 4 and older: 0.20–0.65 Hz; Bar-Haim et al., 2000).

Data Analysis

Data screening and preliminary analyses were conducted in IBM SPSS 24. The three series of multigroup multivariate regression models were conducted in Mplus 8.4. For all models, the multigroup function allowed for examining males and females separately, and the dependent variables were irritability and defiance/vindictiveness. RSA values were natural log transformed prior to analysis. Each multigroup multivariate regression model was run using maximum likelihood estimation. All covariates were brought into the model so observations with missing data were not deleted. The GROUPING command was used to do multiple-group analysis by gender. Physiological variables used in interactions were centered. Bootstrapping (with 10,000 samples) was used to obtain standard errors and non-symmetric confidence intervals.

The first series of models used the trauma type variable and included those with interpersonal trauma and those with non-interpersonal trauma. In the first model trauma type, IBI, and the IBI by trauma type interaction were the independent variables. The second and third models were similar but included RSA and cortisol in place of IBI (see Figure 1 for example). A second series of three models (one for each physiological variable) was run including only those children who experienced interpersonal trauma and controls. Finally, a third series of models was run including only those children who experienced non-interpersonal trauma and controls.

Figure 1.

Visual Representation of Model Using RSA by Trauma Type Interaction

Results

Within the trauma-exposed group, 25.0% (N = 71) had experienced interpersonal trauma only, and 65.1% (N = 185) had experienced non-interpersonal trauma only. The 27 children who had experienced both interpersonal and non-interpersonal trauma were excluded from analyses. Of those who had experienced interpersonal trauma only, 54.9% (N = 39) met criteria for the full ODD diagnosis. Of those who had experienced non-interpersonal trauma only, 28.4% (N = 52) met criteria for ODD. Of all those who met criteria for ODD within the trauma-exposed group, 42.9% had experienced interpersonal trauma only and 57.1% had experienced non-interpersonal trauma only.

Preliminary analyses showed no evidence of multicollinearity among predictors for each model (no tolerance values less than .10, no variance inflation factor (VIF) values over 10, no correlations above .70). Residuals for each model were normally distributed. Several outliers were found with Mahalanobis distances greater than the critical value (two IB I, and three cortisol). These outliers were each recoded to be one unit larger than the most extreme score in their respective distributions, per Tabachnick and Fidell (2013). After recoding outliers, skewness and kurtosis for each variable were within acceptable ranges.

Simple bivariate correlations and descriptive statistics for variables after excluding children with both trauma types, transformations, and adjusting outliers are presented in Table 1. Nonparametric correlations between variables are presented separately by gender, and differences in patterns of significance further justified the decision to examine males and females separately. Interpersonal trauma was significantly correlated with irritability and defiance/vindictiveness among males and females. Non-interpersonal trauma was not related to either irritability or defiance/vindictiveness in males or females. Of the physiological variables, only cortisol was significantly correlated with irritability in females. Age was not significantly related to irritability or defiance/vindictiveness and thus was not included in the models.

Table 1.

Nonparametric Correlations and Descriptive Statistics for Study Variables

	age	IRR	DEFV	IBI	RSA	COR	INT	NINT
Age	--
IRR	−.13 / −.05	--
DEFV	−.14 / .02	.48 / .61	--
IBI	.37 / .34	−.14 / −.04	−.13 / −.07	--
RSA	.39 / .12	−.16 / .01	−.08 / −.07	.75 / .76	--
COR	.15 / −.04	−.24 / −.04	−.11 / −.07	.26 / −.06	.30 / −.00	--
INT	−.08 / −.02	.25 / .25	.20 / .23	−.13 / .10	−.17 / .08	−.09 / −.05	--
NINT	.13 / .14	−.05 / −.00	−.03 / −.00	.10 / −.01	.05 / −.00	.01 /.10	−.70 / −.70	--
Total N	303	301	301	232	232	229	302	302
Mean (SD)	5.01 (1.08)	1.36 (1.19)	1.79 (1.61)	644.43 (92.67)	4.54 (0.74)	8.17 (5.87)	--	--
Skewness (SE)	0.08 (0.14)	0.18 (0.14)	0.48 (0.14)	0.48 (0.16)	−0.34 (0.16)	2.32 (0.16)	--	--
Kurtosis (SE)	−1.08 (0.28)	−1.49 (0.28)	−0.95 (0.28)	0.18 (0.32)	0.07 (0.32)	6.32 (0.32)	--	--

Note. IRR = symptoms of oppositional defiant disorder irritability; DEFV = symptoms of oppositional defiant disorder defiance/vindictiveness; IBI = interbeat interval; RSA = respiratory sinus arrythmia; COR = cortisol; INT = interpersonal trauma; NINT = non-interpersonal trauma.

Correlations (Spearman’s rho) are presented separately by gender: females / males.

Correlations in bold are significant at the p < 0.05 level.

Comparing Interpersonal and Non-interpersonal Trauma (Excluding Controls)

The sample for the first series of models was restricted to only those participants who experienced trauma, excluding participants in the control group. The trauma type variable was used (1 = interpersonal trauma only, and 0 = non-interpersonal trauma only). First, a multiple group analysis (grouping by gender) without the interaction term was run. In Step 1, trauma type and baseline IBI (the inverse of HR) were modeled as predictors of irritability and defiance/vindictiveness. Results showed that trauma type was a positive significant predictor of both irritability and defiance/vindictiveness for males, but only irritability for females (controlling for baseline IBI). Baseline IBI was not a significant predictor of irritability or defiance/vindictiveness for males or females. Then, in Step 2, the IBI by trauma interaction term was added to the model but was not a significant predictor of irritability or defiance/vindictiveness for males or females (Table 2).

Table 2.

Parameter Estimates for Models Including Only Those with Interpersonal Trauma and Non-Interpersonal Trauma, (Excluding Controls N = 256)

	Males				Females
	Standardized estimate	Unstandardized estimate	SE	p	Standardized estimate	Unstandardized estimate	SE	p
IBI model - Step 1
IRR
TRA	0.191	0.501	0.199	.012	0.226	0.589	0.250	.018
IBI	−0.006	0.000	0.001	.950	−0.060	−0.001	0.001	.613
DEFV
TRA	0.181	0.651	0.286	.023	0.140	0.497	0.333	.135
IBI	0.001	0.000	0.002	.994	−0.088	−0.001	0.002	.493
IBI model - Step 2
IRR
IBIxTRA	0.101	0.002	0.003	.421	−0.098	−0.003	0.004	.421
DEFV
IBIxTRA	0.137	0.004	0.004	.288	0.027	0.001	0.005	.816
RSA model - Step 1
IRR
TRA	0.189	0.496	0.197	.012	0.226	0.591	0.251	.019
RSA	0.012	0.019	0.150	.898	−0.061	−0.100	0.183	.585
DEFV
TRA	0.184	0.661	0.287	.021	0.156	0.552	0.335	.100
RSA	−0.027	−0.058	0.200	.772	0.002	0.005	0.291	.985
RSA model - Step 2
IRR
RSAxTRA	0.028	0.096	0.381	.801	−0.186	−0.761	0.497	.126
DEFV
RSAxTRA	0.138	0.657	0.501	.190	−0.065	−0.360	0.722	.618
COR model - Step 1
IRR
TRA	0.188	0.494	0.197	.012	0.222	0.581	0.239	.015
COR	−0.073	−0.014	0.018	.433	−0.099	−0.024	0.030	.433
DEFV
TRA	0.180	0.647	0.288	.025	0.140	0.498	0.344	.147
COR	−0.052	−0.014	0.028	.621	−0.109	−0.035	0.032	.274
COR model - Step 2
IRR
CORxTRA	−0.049	−0.016	0.043	.711	−0.242	−0.208	0.111	.061
DEFV
CORxTRA	−0.058	−0.025	0.066	.703	0.159	0.185	0.122	.130

Note. IRR = symptoms of oppositional defiant disorder irritability; TRA = trauma type (1 = interpersonal trauma, 0 = non-interpersonal trauma); IBI = interbeat interval; IBIxTRA = IBI by trauma type interaction; DEFV = symptoms of oppositional defiant disorder defiance/vindictiveness; RSA = respiratory sinus arrythmia; RSAxTRA = RSA by trauma type interaction; COR = cortisol; CORxTRA = cortisol by trauma type interaction.

Next, a similar set of models was run examining RSA in place of IBI. The initial model showed trauma type was a significant predictor of irritability and defiance/vindictiveness for males, and irritability for females (controlling for baseline RSA). Baseline RSA was not a significant predictor of irritability or defiance/vindictiveness for males or females. The RSA by trauma type interaction term was then added to the model and was not a significant predictor.

Baseline cortisol was then examined. Again, the Step 1 model showed that trauma type was a significant predictor of irritability and defiance/vindictiveness for males, and irritability for females (controlling for baseline cortisol). Baseline cortisol was not a significant predictor of any outcome variables. When the cortisol by trauma type interaction term was added to the model in Step 2, it also was not a significant predictor of any outcome variable but was trending toward significance in the prediction of irritability for girls (see Table 2).

Comparing Interpersonal Trauma and Controls

The sample for the second series of models was restricted to only those participants who experienced interpersonal trauma and non-trauma-exposed (controls). The interpersonal trauma variable was used (1 = interpersonal trauma only, 0 = control). Analyses were conducted in the same order as the interpersonal versus non-interpersonal models above and the results are detailed in Table 3. We highlight here only the differences from the results of the interpersonal versus non-interpersonal models for conciseness. The effect of baseline RSA on irritability was significant for females, suggesting females with lower baseline RSA had higher irritability scores (controlling for interpersonal trauma). Also, the interaction between RSA and interpersonal trauma was a significant predictor of defiance/vindictiveness for males only (Table 3; R-square = 0.291, SE = 0.090, p = .001). Effect plots (not included) showed that among males, interpersonal trauma has a larger positive effect on defiance/vindictiveness with increasing values of RSA, and that effect was significant in the centered RSA range of −0.6 and above. Figure 2 shows the significant interaction predicting defiance/vindictiveness in males, as well as the non-significant interaction predicting defiance/vindictiveness in females for comparison.

Table 3.

Parameter Estimates for Models Including Only Those with Interpersonal Trauma and Controls (N = 117)

	Males				Females
	Standardized estimate	Unstandardized estimate	SE	p	Standardized estimate	Unstandardized estimate	SE	p
IBI model – Step 1
IRR
INT	0.559	1.342	0.221	<.001	0.428	1.047	0.393	.008
IBI	−0.057	−0.001	0.002	.713	−0.226	−0.003	0.002	.120
DEFV
INT	0.479	1.651	0.328	<.001	0.359	1.122	0.529	.034
IBI	−0.005	0.000	0.003	.974	−0.059	−0.001	0.003	.758
IBI model – Step 2
IRR
IBIxINT	0.376	0.006	0.003	.055	−0.024	−0.001	0.005	.919
DEFV
IBIxINT	0.407	0.009	0.005	.061	0.006	0.000	0.008	.984
RSA model – Step 1
IRR
INT	0.563	1.350	0.218	<.001	0.338	0.826	0.410	.044
RSA	−0.097	−0.148	0.200	.460	−0.369	−0.803	0.326	.014
DEFV
INT	0.485	1.634	0.324	<.001	0.340	1.063	0.588	.071
RSA	−0.038	−0.081	0.282	.773	−0.084	−0.234	0.546	.669
RSA model – Step 2
IRR
RSAxINT	0.177	0.402	0.391	.304	0.024	0.067	0.781	.932
DEFV
RSAxINT	0.338	1.082	0.509	.033	−0.155	−0.564	1.388	.685
COR model – Step 1
IRR
INT	0.546	1.310	0.227	<.001	0.413	1.009	0.370	.006
COR	−0.006	−0.001	0.022	.962	−0.266	−0.058	0.048	.224
DEFV
INT	0.479	1.616	0.324	<.001	0.383	1.198	0.508	.018
COR	−0.031	−0.008	0.029	.792	0.041	0.012	0.058	.842
COR model – Step 2
IRR
CORxINT	−0.312	−0.067	0.047	.155	−0.291	−0.177	0.124	.153
DEFV
CORxINT	−0.212	−0.064	0.073	.384	0.149	0.116	0.144	.422

Note. IRR = symptoms of oppositional defiant disorder irritability; INT = interpersonal trauma (1 = interpersonal trauma, 0 = no trauma control); IBI = interbeat interval; IBIxINT = IBI by interpersonal trauma interaction; DEFV = symptoms of oppositional defiant disorder defiance/vindictiveness; RSA = respiratory sinus arrythmia; RSAxINT = RSA by interpersonal trauma interaction; COR = cortisol; CORxINT = cortisol by interpersonal trauma interaction.

Figure 2.

Baseline RSA by interpersonal trauma interaction predicting defiance/vindictiveness

Also, the results showed a significant positive effect of interpersonal trauma on defiance/vindictiveness for females (in the IBI and COR models, and trending in the RSA model) that was not found in the earlier comparison of interpersonal versus non-interpersonal trauma.

Comparing Non-interpersonal Trauma and Controls

The sample for this series of models was restricted to only those participants who experienced non-interpersonal trauma and controls. The non-interpersonal trauma variable was used (1 = non-interpersonal trauma only, 0 = control). Analyses were conducted in the same order as the models above and the results are detailed in Table 4. We highlight here only the differences from the results of the interpersonal versus non-interpersonal models for conciseness. Non-interpersonal trauma was significantly associated with irritability and defiance/vindictiveness for males, but not females.

Table 4.

Parameter Estimates for Models Including Only Those with Non-Interpersonal Trauma and Controls (N = 231)

	Males				Females
	Standardized estimate	Unstandardized estimate	SE	p	Standardized estimate	Unstandardized estimate	SE	p
IBI model – Step 1
IRR
NINT	0.286	0.838	0.188	<.001	0.187	0.559	0.314	.075
IBI	−0.113	−0.002	0.001	.234	−0.071	−0.001	0.001	.553
DEFV
NINT	0.258	0.998	0.255	<.001	0.149	0.612	0.433	.158
IBI	−0.120	−0.002	0.002	.193	−0.090	−0.001	0.002	.485
IBI model – Step 2
IRR
IBIxNINT	0.250	0.004	0.002	.094	0.201	0.003	0.004	.481
DEFV
IBIxNINT	0.229	0.005	0.003	.144	−0.052	−0.001	0.007	.894
RSA model – Step 1
IRR
NINT	0.281	0.823	0.191	<.001	0.184	0.550	0.317	.083
RSA	−0.056	−0.085	0.144	.556	−0.036	−0.061	0.191	.751
DEFV
NINT	0.259	1.001	0.255	<.001	0.153	0.627	0.446	.159
RSA	−0.135	−0.271	0.185	.142	0.033	0.076	0.312	.808
RSA model – Step 2
IRR
RSAxNINT	0.204	0.356	0.255	.163	0.502	0.890	0.656	.175
DEFV
RSAxNINT	0.168	0.385	0.313	.219	−0.074	−0.179	1.214	.883
COR model – Step 1
IRR
NINT	0.275	0.807	0.197	<.001	0.179	0.534	0.317	.092
COR	0.015	0.003	0.022	.888	−0.098	−0.020	0.022	.369
DEFV
NINT	0.247	0.955	0.260	<.001	0.140	0.572	0.446	.200
COR	0.017	0.005	0.031	.883	−0.116	−0.032	0.026	.221
COR model – Step 2
IRR
CORxNINT	−0.213	−0.050	0.042	.229	0.114	0.029	0.058	.621
DEFV
CORxNINT	−0.133	−0.041	0.060	.494	−0.136	−0.047	0.088	.593

Note. IRR = symptoms of oppositional defiant disorder irritability; NINT = non-interpersonal trauma (1 = non-interpersonal trauma, 0 = no trauma control); IBI = interbeat interval; IBIxNINT = IBI by non-interpersonal trauma interaction; DEFV = symptoms of oppositional defiant disorder defiance/vindictiveness; RSA = respiratory sinus arrythmia; RSAxNINT = RSA by non-interpersonal trauma interaction; COR = cortisol; CORxNINT = cortisol by non-interpersonal trauma interaction

Pre-trauma ODD

The relationship between pre-trauma ODD symptoms and type of trauma was examined first. Of the 254 children who experienced either interpersonal or non-interpersonal trauma (excluding those with both and those missing ODD data), 105 children showed the first onset of ODD symptoms prior to trauma exposure at age three years or older. Of the children who experienced interpersonal trauma only, 45.1% had an onset of ODD symptoms prior to trauma, and 39.9% of children who experienced non-interpersonal trauma only had an onset of ODD symptoms prior to trauma. The association between type of trauma and ODD symptom onset prior to the earliest trauma was not significant (chi-square with continuity correction = 0.373, df = 1, p = .542).

Next, the relationship between the full ODD diagnosis and type of trauma was examined. Of the 254 children who experienced either interpersonal or non-interpersonal trauma (excluding those who experienced both and those missing ODD data), 44 children who met criteria for ODD diagnosis began exhibiting ODD symptoms prior to trauma exposure at age three years or older. Of children who experienced interpersonal trauma only, 28.2% who met criteria for ODD diagnosis began exhibiting ODD symptoms prior to their earliest trauma, compared to 13.1 % of children who experienced non-interpersonal trauma only. There was a significant association between type of trauma and ODD diagnosis onset prior to earliest trauma (chi-square with continuity correction = 7.077, df = 1, p = .008, phi = −0.179).

Discussion

The goals of the current study were to better understand the mixed results presented in the literature regarding the relationships between baseline physiological arousal, trauma type, and symptoms of ODD. It was expected that in the context of interpersonal trauma, low physiological arousal (low HR/high IBI, low RSA, low cortisol) would be related to high ODD (particularly defiant/vindictive) symptoms. If this hypothesis was supported, it would potentially help to explain the inconsistent relationships found between physiological variables and ODD. However, results from the current study did not support this hypothesis and instead provided preliminary findings of a more complicated picture. We believe this study contributes four important findings.

First, the test of an interaction effect between baseline physiological arousal variables and trauma type in predicting ODD dimensions was not supported in the primary comparison of interpersonal versus non-interpersonal trauma. The interaction effect was not found when testing three different physiological variables – IBI, RSA, and cortisol – and grouping by sex. However, the availability of an age-matched non-trauma-exposed control group allowed additional analyses. The baseline RSA by interpersonal trauma interaction was a significant predictor of defiance/vindictiveness among boys, but not girls, when interpersonal trauma was compared to controls. Boys with interpersonal trauma and high RSA had the highest defiance/vindictiveness symptoms. This finding contradicts the hypothesis of the study, which predicted that in the context of interpersonal trauma, lower RSA would be related to higher defiance/vindictiveness. Because trauma wasn’t assessed or reported on in several prior studies reporting a relationship between lower RSA and higher externalizing behaviors (Beauchaine, 2001; Beauchaine et al., 2007; Beauchaine et al., 2013), it is difficult to know how the current results compare. Additional studies specifically including traumatized and non-traumatized individuals are needed before conclusions can be drawn.

Second, except for the single significant effect for the interaction between RSA and interpersonal trauma when restricted to males and the comparison of interpersonal trauma to controls, baseline physiology demonstrated little explanatory power in predicting ODD symptoms. When simple bivariate correlations between baseline physiological variables and ODD dimensions were examined, the only significant correlation was a negative relationship between baseline cortisol and irritability among females. While consistent with the overall results of the meta-analysis by Alink and colleagues (2008), these results are inconsistent with moderator analyses in the same study which showed that preschoolers with higher baseline cortisol had higher externalizing behaviors. The current findings are also inconsistent with previous studies reporting gender differences which showed lower cortisol was related to higher externalizing in boys, but not girls (Dorn et al., 2009; Shirtcliff et al., 2005). It should be noted that only a single baseline cortisol measurement was used in the current study, while others have used multiple samples to establish baseline levels (Dorn et al., 2009; Shirtcliff et al., 2005). When trauma was controlled for in multiple regression models in the current study, baseline cortisol was no longer a significant predictor of irritability among females.

In addition to the baseline cortisol findings, baseline RSA was a significant negative predictor of irritability for females (controlling for trauma) when the sample was restricted to those who experienced interpersonal trauma and controls. This is consistent with some previous research (Beauchaine et al., 2013), though inconsistent with other studies that have specifically examined sex differences (Beauchaine et al., 2008).

There may be several explanations for why we generally failed to find associations with physiological variables in contrast to prior studies. Foremost may be that there are no other known studies that have examined all three constructs of true trauma exposure, baseline physiological arousal, and ODD in children simultaneously. The one other known study which has examined baseline RSA in relation to ODD as a multidimensional construct across gender found no relationship between baseline RSA and irritable or defiant/vindictive symptoms for males or females (Vidal-Ribas et al., 2017). Another reason why others may have failed to find a relationship between baseline RSA and ODD is that there may be true gender differences, as others have suggested (Hinnant & El-Sheikh, 2013), which attenuate findings when they are not accounted for. While there is no evidence that the stress response system itself functions differently across sex, it may be the case that the behavioral manifestations associated with physiological arousal differ across gender. Last, almost all of the prior studies in children examined older populations.

Third, the relationships between trauma type and ODD dimensions revealed gender differences. When comparing interpersonal with non-interpersonal trauma or controls, interpersonal trauma was significantly associated with higher irritability symptoms within females. Non-interpersonal trauma was not significantly related to irritability when compared to controls. In addition, when comparing interpersonal with non-interpersonal trauma within females, interpersonal trauma was not predictive of defiance/vindictive symptoms. However, when interpersonal trauma was compared to controls, interpersonal trauma was significantly associated with higher defiance/vindictive symptoms. The totality of these gender findings provides preliminary support for analyzing ODD as two dimensions rather than a unitary construct, at least for females. In contrast, both interpersonal and non-interpersonal trauma were predictive of both irritability and defiance/vindictiveness within males.

Fourth, the findings discussed suggest the existence of ODD symptoms is different for those who have experienced interpersonal trauma compared to those who have not. It may be tempting to assume that the experience of trauma itself leads to those differences. However, the results discussed so far come from cross-sectional assessments which possess no causal explanatory power and it is impossible to know which of the three variables (individual differences in physiological arousal, ODD symptoms, or trauma) occurred first. It may be that the children who experienced interpersonal trauma had higher RSA and ODD symptoms prior to that trauma. For example, at least one study showed that ODD was associated with risk for later trauma (Carliner et al., 2017).

The current study addressed that gap partially by asking caregivers to retrospectively determine if the onset of ODD symptoms preceded or followed the first trauma exposures. While the potential biases of retrospective recall are acknowledged, the potential for bias for the lifespan of such young children are less than that for adults. Results showed that a larger percentage of children who met criteria for ODD diagnosis and experienced interpersonal trauma began exhibiting ODD symptoms prior to their trauma compared to those who experienced non-interpersonal trauma. These data provide some evidence that children who experience interpersonal trauma may be different from those who experience non-interpersonal trauma before those traumas occur. This idea is consistent with findings by Danese (2020) and suggests that those who experience interpersonal trauma may have pre-existing vulnerabilities that make them more likely to both exhibit ODD symptoms and experience trauma. Based on the current study data, one of those pre-existing vulnerabilities for boys may be high resting RSA. However, because RSA was only collected at one time point, this hypothesis could not be tested in the current data. As alluded to, a limitation to the current study and most of the research in this area is the cross-sectional nature of it. There are obvious practical and ethical reasons why it is difficult to prospectively measure physiology and psychopathology prior to trauma exposure in children. Nonetheless, without longitudinal data, causal conclusions cannot be made.

Another limitation to the current study is that each psychophysiological variable was examined separately, and the joint activation of the PNS and SNS could not be examined. Others have suggested that moving beyond single biomarkers and examining multisystem functioning is an important next step in psychophysiological research (Buss et al., 2018). While the PNS and SNS often have opposing effects, their responses are not necessarily coupled (Berntson et al., 1991). Whether both systems are co-activated, co-inhibited, or reciprocally activated may provide more information than the functioning of a single system. For example, El-Sheikh et al. (2009) showed that in the context of marital conflict, PNS and SNS co-activation and co-inhibition served as vulnerability factors for externalizing behavior, whereas reciprocal activation served as a protective factor. While RSA provides a measure of PNS functioning, the current study did not include a measure of SNS functioning, such as pre-ejection period or skin conductance level. Further, only baseline levels of physiological arousal were analyzed in the current study, whereas examining responding to a relevant stressor would also provide additional information about ANS functioning.

The current study is also limited in terms of the generalizability of results. Specifically, while the racial composition of the sample was generally representative of the population where the study was conducted, a majority of the sample identified as Black. This may limit the generalizability of the findings to other populations. Further, given that the focus of the study was on examining the relationship between types of trauma and ODD, including a traumatized sample was necessary. However, the sample of children included in this study are not necessarily representative of the general population of children with ODD.

Aside from excluding children who experienced both interpersonal and non-interpersonal trauma, the current study did not take number of traumas into account. It is likely, given the nature of interpersonal trauma (including witnessing violence and experiencing abuse), that a greater proportion of children in the interpersonal trauma group experienced repeated traumas compared to the non-interpersonal group. Relatedly, it is acknowledged that children exposed to interpersonal trauma are more likely to be exposed to coercive parenting practices and other family-based adversities. This highlights a significant limitation that impacts every study that groups individuals by interpersonal versus non-interpersonal trauma. The principle of exchangeability in cross-sectional research states that groups need to be equal on all salient variables prior to their exposures to the variable of interest. When groups are not equal on all salient variables (i.e., interpersonal trauma victims experience additional family-based adversities and non-interpersonal trauma victims do not), this creates a challenge for controlling all salient variables and for interpreting the results. The potential effects of parenting practices could not be controlled for in this study. The main implication of this is that the effects of family adversities, if those existed, could not be parsed from the effects of interpersonal trauma exposure on ODD symptoms. Finally, there are many other ways of grouping trauma beyond the interpersonal/non-interpersonal distinction, such as deprivation- vs. threat-based, level of harshness and unpredictability, and familial vs. extra-familial (Lacey et al., 2020) which may have differing relationships with ODD.

In conclusion, the current study was the first to examine the relationship between true trauma exposure, baseline physiological arousal, and their interaction in predicting dimensions of ODD across gender. Although several hypotheses were not supported, the current study highlighted several avenues for future research. Namely, the current findings differed based on gender, trauma exposure, and ODD dimensions, which may begin to explain the mixed results found in previous literature, and provide a stepping stone for more nuanced hypotheses moving forward. We also provided preliminary evidence that children who will one day experience interpersonal trauma are different from those who will experience non-interpersonal trauma in regard to pre-existing ODD. In other words, most victims of interpersonal and non-interpersonal trauma are different prior to any trauma exposure, violating the principle of exchangeability in cross-sectional research, and emphasizes a point that conclusions about interpersonal trauma drawn from cross-sectional studies require confirmation in prospective longitudinal studies. These findings in the context of prior studies highlight some of the difficulties of conducting research with trauma populations that need to be considered in future studies.