Neural correlates of the propensity for retaliatory behavior in youth with disruptive behavior disorders

Abstract

Objective

Youth with Disruptive Behavior Disorders (DBD: Conduct Disorder/Oppositional Defiant Disorder) are at increased risk for maladaptive reactive aggression. Theory suggests this is due to increased sensitivity of basic threat circuitry implicated in retaliation (amygdala/periaqueductal gray) in youth with DBD and low levels of Callous-Unemotional Traits and dysfunctional regulatory activity within ventromedial prefrontal cortex (vmPFC) in youth with DBD irrespective of callous-unemotional traits.

Methods

Fifty-six youths participated (23 female) aged 10–18 (26 healthy and 30 with DBD [15 with high and 15 with low callous-unemotional traits]) who completed an Ultimatum Game during functional MRI.

Results

Youth with DBD and low callous-unemotional traits showed greater increases in activation of basic threat circuitry when punishing others relative to comparison groups and dysfunctional down regulation of vmPFC during retaliation. All youth with DBD showed reduced amygdala-vmPFC connectivity during high provocation relative to healthy youth. VmPFC responsiveness and vmPFC-amygdala connectivity were related to patients’ retaliatory propensity (behavioral responses during task) and parent reported reactive aggression.

Conclusions

These data suggest differences in the underlying neurobiology of maladaptive reactive aggression in youth with DBD and low relative to high levels of callous-unemotional traits. Youth with DBD and low callous-unemotional traits alone show significantly greater threat responses during retaliation relative to comparison individuals. Moreover, these data suggest that vmPFC-amygdala connectivity is critical for regulating retaliation/reactive aggression and when dysfunctional, contributes to reactive aggression, independent of level of callous-unemotional traits.

The term “reactive aggression” has been used to describe retaliatory actions in response to perceived threat, frustration, or provocation.¹ Youth with Disruptive Behavior Disorders (DBD), such as Conduct Disorder and Oppositional Defiant Disorder commonly display mal-adaptively higher levels of reactive aggression than youth without psychopathology.^2,3 Such high levels of aggression typically occur in the context of other forms of antisocial behavior,⁴ and decision-making deficits.⁵ The current study examines the neural correlates of retaliatory actions in healthy youth and youth with DBD.

Prior research examines such neural correlates based on the response to social provocation in variants of the Ultimatum Game (UG).⁶ In one variant, a computer-simulated partner proposes to divide resources with participants. Participants accept or reject the division. Participants are more likely to reject an unfair offer if the partner could have offered a fair allocation, but the magnitude of this effect is reduced in delinquents/offenders.^7,8 Another UG variant involves participants choosing whether to “punish” the partner by spending money to remove the partner’s money. Such behaviors expressed in this variant model retaliatory behavior in participants. Notably, increasing levels of retaliatory behavior are associated with increased activity in regions implicated in reactive aggression in animal studies (amygdala, periaqueductal gray [PAG]).^9–12 Furthermore, neuroimaging studies implicate an inverse relationship between activation in ventromedial prefrontal cortex (vmPFC) and amygdala/PAG response to basic threats¹³ and provocation.^9,14 VmPFC is thought to be critical for representing expected value (i.e. the subjective reward value associated by an individual with an action/object following learning¹⁵). Decreased vmPFC activation may reflect the representation of the diminishing expected value of increasing levels of retaliation, as increasing retaliation leads to diminishing monetary gains in the UG.^9,14 In threat induction paradigms, the increasing proximity of threat is associated with a diminishing likelihood of escape (and increased likelihood of punishment); the value of future behavioral options diminishes as threat increases in proximity.¹³ We have suggested that the increased risk for reactive aggression in patients with DBD reflects increased activity in threat response regions and/or dysfunction in vmPFC modulatory activity.¹⁶

However, youth with DBD are heterogeneous. Some show elevated levels of callous-unemotional (DBD+CU) traits (i.e., decreased guilt/empathy – referred to as “with limited prosocial emotions” in DSM-5), while others do not (DBD−CU). Youth with DBD+CU and DBD−CU have distinct developmental trajectories and neurobiology.^16,17 Both groups, however, show elevated levels of reactive aggression.³ The neurobiology of reactive aggression might be different for each group. We have hypothesized that patients with DBD−CU show increased risk for reactive aggression because of heightened responsiveness of basic threat systems (amygdala/PAG); the suggestion is that this increased responsiveness increases the probability that a given level of provocation will initiate fighting rather than flight or freezing.¹⁶ There is evidence that youth with DBD−CU show increased, while youth with DBD+CU show decreased, amygdala responses to fearful expressions.^18–20 Alternatively, there may be common neurobiological underpinnings of reactive aggression, which we have hypothesized reflect dysfunction in vmPFC’s modulatory role in representing an action’s expected value.¹⁶ Expected value representation is disrupted in patients with DBD irrespective of CU levels.^5,21 In the context of a UG, we expect that youth with DBD will fail to appropriately represent the diminishing value of outcomes as retaliation increases, leading to increased propensity to retaliate. Here, we examine these possibilities through investigation of retaliatory behavior and its neural underpinnings in the context of a UG.

This study tested four predictions: First, retaliatory propensity, measured by average levels of punishment during the UG, would be significantly more associated with reactive than proactive aggression. Second, the increase in amygdala/PAG activity when retaliating would be exaggerated in youth with DBD−CU relative to healthy youth and the level of activity in these regions would be positively associated with reactive aggression. Third, the typical decrease in vmPFC activity when retaliating and vmPFC-amygdala functional connectivity would both be attenuated in youth with DBD−CU and DBD+CU relative to healthy youth. Fourth, the level of attenuation of modulated response and the connectivity during high provocation/retaliation in vmPFC would be negatively associated with retaliatory propensity.

Methods

Participants

Fifty-six youths (23 female) participated: 30 youth with DBD and 26 healthy youth aged 10 to 18 [mean=14.74 (2.13)]. Consistent with previous work,¹⁸ a median split (median=43.5) approach was utilized to form DBD+CU [mean=46.8 (3.57)] and DBD−CU [mean=33.93 (6.375)] groups. Groups did not significantly differ on age, IQ, or in gender composition (Table 1). Half of the youth with DBD (n=15) were comorbid for Attention Deficit-Hyperactivity Disorder and medications could not be withheld during scanning for 23.3% (n=7) of the DBD group. Youth were recruited from the community. After complete description of the study to the subjects, written informed assent/consent was obtained from children and parents. The NIH IRB approved this study. See Supplemental Material 1.1 for inclusion/exclusion criteria. Groups did not differ on movement during scanning [t=.45, p=.66].

Table 1.

Demographics and task performance on the Social Fairness Game in 26 healthy youth and 30 youth with disruptive behavior disorders.

	Health Youth		Youth with DBD
	mean	std. dev.	mean	std. dev.	t	p
age	14.30	2.19	15.04	2.09	1.13	.26
IQ	98.96	9.99	96.37	10.52	.94	.35
reactive aggression	0.85	0.97	3.97	1.43	9.43	<.01
proactive aggression	0.08	0.27	2.57	2.01	6.25	<.01
	N	percentage	N	percentage	χ2	p
female participants	12	46.2%	11	36.7%	0.52	.47
youth medicated during scanning	0	0%	7	23.3%	6.93	<.01
youth with ADHD	0	0%	15	50.0%	17.76	<.01
youth with conduct disorder	0	0%	22	73.3%	9.29	<.01
youth with ODD			9	30.0%	32.87	<.01
	Punishment dollars spent		Punishment dollars spent
Offer Unfairness Level	mean	std. dev.	mean	std. dev.	t	p
$10/$10	0.17	0.27	0.19	0.35	0.30	.77
$14/$6	1.17	0.71	1.56	0.78	1.98	.05
$16/$4	1.63	0.70	1.86	0.66	1.25	.22
$18/$2	2.18	0.62	2.25	0.70	0.36	.72
	Response Latency		Response Latency
	mean	std. dev.	mean	std. dev.	t	p
$10/$10	918.81	235.12	931.59	297.96	0.18	.86
$14/$6	1121.67	289.50	1174.56	475.17	0.49	.62
$16/$4	1179.67	293.67	1214.88	452.25	0.34	.74
$18/$2	1019.54	333.67	1112.19	452.37	0.86	.39

std. dev = standard deviation, DBD= Disruptive Behavior Disorder, ADHD= Attention Deficit-Hyperactivity Disorder, ODD= Oppositional Defiant Disorder

Study Measures

Inventory of Callous-Unemotional Traits.²²

The ICU is a 24-item parent-report scale designed to assess CU traits in youth. The ICU was derived from the CU scale of the Antisocial Process Screening Device²³ that has been used in various youth samples. The construct validity of the ICU has been supported in community and juvenile justice samples.^24,25

Reactive/Proactive Aggression Rating Scale

Parents completed this six item measure.¹ Dodge and Coie found that a 3 item reactive aggression scale and a 3 item proactive aggression scale yielded the best fitting model¹ and was found to be useful in distinguishing aggression sub-types in children¹ and adolescents.²⁶

The Social Fairness Game

Participants were presented with a version of the Ultimatum game, the Social Fairness game, that has been previously described;¹⁴ see Figure 1 and Supplemental Materials 1.2. Participants are offered either a fair ($10 to participant; $10 to partner) or unfair ($6/$4/$2 to participant; $14/$16/$18 to partner) division of a $20 pot. Participants could then either accept the offer or reject it and, by spending $1, $2 or $3 punishment dollars, punish the partner. Each punishment dollar spent by the participant caused the partner to lose $7 from the $20 pot (Figure 1). Retaliatory propensity was defined as the participant’s average punishment level.

Figure 1. The Social Fairness Game.

Participants were presented with either a fair (A) or unfair (B/C/D) division $20. After a jittered interval, participants chose, via button press, either to accept the offer (E) or to retaliate by spending $1–$3 of a pot of punishment money. For every punishment dollar spent in retaliation, the offerer lost $7 (F/G/H).

MRI parameters and preprocessing

Participants were scanned using a 3T GE scanner and data were analyzed using Analysis of Functional Neuroimages (AFNI).²⁷ Specific parameters have been reported elsewhere¹⁴ and can be found in Supplemental Materials 1.3.

General Linear Model (GLM) Analysis

The model involved six motion regressors and the following task regressors: indicator functions for the (i) offer-phase; (ii) decision-phase; (iii) outcome-phase; (iv) offer-phase multiplied by offer unfairness (0 [fair: $10 kept; $10 offered to partner], 8 [unfair: $14 kept; $6 offered to partner], 12 [unfair: $16/$4], 16 [unfair: $18/$2]); and (v) decision-phase multiplied by the sample average punishment level at each level of unfairness (average response to $10/$10=1.176, $14/$6=2.358, $16/$4=2.726, $18/$2=3.23). All regressors were convolved with a canonical hemodynamic response function from the onset of the offer-, decision- and outcome-phases to account for the slow hemodynamic response. In accordance with findings that normalization of brain volumes from age 7–8 years onward does not introduce major age-related distortions in localization or time course of the blood-oxygen-level-dependent (BOLD) signal in event-related fMRI,^28,29 the participants’ anatomical scans were individually registered to the Talairach and Tournoux atlas.³⁰ The individuals’ functional echo planar image data were then registered to their Talairach anatomical scan. Linear regression modeling was performed using the 5 regressors described earlier, plus regressors to model a first-order baseline drift function. This produced β coefficients and associated t statistics for each voxel and regressor.

fMRI data analysis

A 3 (diagnosis: DBD+CU, DBD−CU, healthy youth) by 2 (task phase: offer-phase, decision-phase) repeated-measures analysis-of-variance (ANOVA) was conducted on the modulated regression coefficients. The AFNI ClustSim program was used to establish a family-wise error corrected threshold (594mm³ clusters at p=.005, corrected to p=.05) for a whole-brain analysis. Due to their small size and/or theoretical importance, small volume corrections (SVC) for multiple comparisons were calculated for amygdala, PAG and vmPFC. The amygdala SVC, calculated using an anatomically defined mask (Eickhoff-Zilles Architectonic Atlas: 50% probability),³¹ yielded a threshold of 162mm³ at an initial threshold of p=.02. As anatomically defined masks were not available in AFNI, the PAG and vmPFC SVCs were calculated using10mm spheres centered on the peak coordinates from previous work (PAG x,y,z= 3,−23,−4; vmPFC x,y,z= −4,36,−5)¹³ and yielded a threshold of 248mm³ at an initial threshold of p=.02 for both regions. For these analyses, average percent signal change was measured across all voxels within each region of interest (ROI) generated from the functional masks and post-hoc testing was conducted in SPSS.³²

Generalized psychophysiological interaction analyses³³ were conducted to investigate differences in functional connectivity associated with DBD during the UG. As a parametric modulation strategy is incompatible with functional connectivity analyses, functional connectivity was compared using two 3 (diagnosis: youth with DBD+CU, youth with DBD−CU, healthy youth) by 2 (provocation level: low [fair/accept], high [unfair/punished]) by 2 (task phase: offer-phase, decision-phase) repeated-measures ANOVAs. The anatomically defined amygdala masks mentioned above were seed regions. See Supplementary Materials 1.4 for details.

Results

Behavioral Results

Prediction 1: Retaliatory propensity, measured by average levels of punishment during the UG, would be significantly more associated with reactive than proactive aggression. Consistent with our prediction, the relationship between retaliatory propensity (average punishment on the UG) and reactive aggression was significantly stronger than the relationship between retaliatory propensity and proactive aggression [Whole group: r=.26, .04 respectively; Steiger’s Z=2.13, p=.03; patients with DBD only: r=.37, −.13 respectively; Steiger’s Z=2.62, p<.01]. Additionally, within youth with DBD, there was a significantly stronger relationship between speed to retaliate and reactive aggression relative to proactive aggression [r=−.59, −.25 respectively; Steiger’s Z=2.00, p<.05]. CU traits within the youth with DBD were significantly related to proactive [r=.45, p=.01], but not reactive aggression [r=.19, p=.33]; though these correlations were not significantly different [Steiger’s Z=1.44, p=.15].

A 4 (offer unfairness: $10/$10, $14/$6, $16/$4, $18/$2) × 2 (diagnosis: youth with DBD, healthy youth) repeated-measures ANOVA was conducted on the choice data (i.e. level of punishment selected: $0, $1, $2, $3). A significant main effect of offer was observed [F(3,54)=420.93, p<.01]. Participants increased retaliation as offer unfairness increased. Neither the main effect of diagnosis [F(3.54)=1.88, p=.18] nor the unfairness level-by-diagnosis interaction was significant [F(3.54)=2.00, p=.12]. However, a significant quadratic unfairness level-by-diagnosis interaction was observed [F(3,54)=4.51, p=.04]. While healthy youth and DBD youth were equally likely to accept fair offers and punished very unfair ($16/$4 and $18/$2) offers equally harshly, DBD youth punished slightly unfair offers ($14/$6) more harshly than healthy youth (Table 1).

fMRI Results

A 3 (diagnosis: DBD+CU, DBD−CU, healthy youth) by 2 (task phase: offer-phase, decision-phase) repeated-measures ANOVA was conducted on the BOLD data modulated by offer unfairness in the offer-phase and punishment level in the decision-phase (for complete results see the Supplemental Materials 2.1–2.2 and Supplementary Tables 1–3).

Prediction 2: The increase in amygdala/PAG activity when retaliating would be exaggerated in youth with DBD−CU relative to healthy youth and the level of activity in these regions would be positively associated with reactive aggression.

ROI Results

Amygdala

A significant main effect of diagnosis was observed (See Figure 2). Youth with DBD−CU showed significantly greater right amygdala (xyz=23,−8,−6, k=10) responses relative to healthy youth [t=3.33, p<.01] and youth with DBD+CU [t=2.84, p<.01] who did not significantly differ [t=.27, p=.79]. Within the youth with DBD, modulated amygdala response was inversely associated with CU traits [r=−.49, p<.01], but not reactive aggression [r=.05, p=.78].

Figure 2. Regions of basic threat circuitry showing group differences between 28 healthy youth, 15 youth with DBD+CU and 15 youth with DBD−CU.

A. During the task, healthy youth show increased modulated activation in right amygdala relative to youth with DBD+CU and DBD−CU.

B. During the decision-, but not the offer-, phase, youth with DBD−CU show increased activation as a function of retaliation in periaqueductal gray relative to youth with DBD+CU and healthy youth.

Y w/ = Youth with

DBD−CU= Disruptive Behavior Disorder with low levels of Callous-Unemotional traits.

DBD+CU= Disruptive Behavior Disorder with high levels of Callous-Unemotional traits.

Periaqueductal Gray

A significant diagnosis-by-task phase interaction was observed in PAG (xyz=10,−21,4, k=6). During the decision phase, youth with DBD−CU showed a significantly greater modulated PAG activation relative to healthy youth [t=2.14, p=.04] and youth with DBD+CU [t=2.23, p=.03] who did not significantly differ [t=.59, p=.56]. There were no significant between-group differences in the offer-phase [t<.94, p>.36]. See Figure 2.

Prediction 3: The typical decrease in vmPFC activity when retaliating and connectivity between the amygdala and vmPFC would both be attenuated in youth with DBD−CU and DBD+CU relative to healthy youth.

vmPFC

A significant main effect of diagnosis was observed where youth with DBD−CU showed significantly reduced attenuation of modulated vmPFC (xyz=−11,33,−2, k=13) response relative to healthy youth [t=3.228, p=.003]. Youth with DBD+CU did not significantly differ from youth with DBD−CU [t=1.64, p=.11] or healthy [t=1.53, p=.14]. See Figure 3.

Figure 3. Group differences between 28 healthy youth, 15 youth with DBD+CU and 15 youth with DBD−CU in vmPFC.

A. Youth with DBD−CU show less reduction in vmPFC activation as a function of retaliation relative to healthy youth and youth with DBD−CU.

B. Healthy youth showed increased functional connectivity during high provocation trials relative to youth with DBD−CU and DBD+CU.

C. A mediation analysis indicated that attenuated modulated vmPFC activation mediated the relationship between reactive aggression and retaliatory propensity.

D. A mediation analysis indicated that vmPFC-amygdala functional connectivity mediated the relationship between reactive aggression and retaliatory propensity.

vmPFC= ventromedial prefrontal cortex

Y w/ = Youth with

DBD−CU= Disruptive Behavior Disorder with low levels of Callous-Unemotional traits.

DBD+CU= Disruptive Behavior Disorder with high levels of Callous-Unemotional traits.

Generalized Psychophysiological Interaction Analyses

A significant diagnosis-by-provocation level interaction was observed in vmPFC using the right amygdala seed. See Figure 3 and Supplemental Materials 2.3. During high provocation conditions, greater functional connectivity was observed between right amygdala and vmPFC in healthy youth relative to both youth with DBD−CU [t=2.98, p<.01] and DBD+CU [t=2.16, p=.04] who did not significantly differ [t=1.26, p=.22].

Prediction 4: The level of attenuation of modulated response and the connectivity in vmPFC during high provocation/retaliation would be negatively associated with retaliatory propensity.

BOLD response and Behavior

Retaliatory propensity was inversely associated with attenuation of modulated vmPFC response [r=−.44, p<.01] and vmPFC-amygdala connectivity during high provocation/retaliation conditions [r=−.34, p=.01]. Given the significant relationship between retaliatory propensity and both modulated vmPFC response and vmPFC-amygdala functional connectivity, a hierarchical linear regression analysis was used to determine whether the modulated and connectivity data were independently related to retaliatory propensity beyond reactive aggression and CU traits. Both modulated BOLD response [β=.36, ΔR²=.11, p<.01] and functional connectivity in vmPFC [β=−.29, ΔR²=.09, p=.02] were observed to contribute unique variance in the prediction of retaliatory propensity.

Attenuation of modulated vmPFC response and retaliatory propensity were both significantly correlated with reactive aggression [r=−.39 & 0.26 respectively, p<.05], though vmPFC-amygdala connectivity was correlated with reactive aggression only at trend levels [r=−.24, p=.08]. The relationship between retaliatory propensity and reactive aggression was not significant when controlling for attenuation in modulated vmPFC response [r_partial=.12, p=.42] or vmPFC-amygdala connectivity [r_partial=.20, p=.15]. However, the relationships between retaliatory propensity and attenuated vmPFC response [r_partial=−.38, p<.01] and vmPFC-amygdala connectivity [r_partial=−.29, p=.03] remained significant controlling for reactive aggression and CU traits. This suggests that modulated vmPFC response and vmPFC-amygdala connectivity mediate³⁴ the association between reactive aggression and retaliatory propensity.

Confounding Factors

Of the youth with DBD, medication could not be withheld during scanning for 23.3% (n=7) and 50% (n=15) were comorbid for ADHD. Therefore, the preceding analyses were re-run removing these youth from the sample. In all regions, activations in similar regions were observed when only un-medicated youth were included and when youth with comorbid ADHD and DBD were excluded (see Supplemental Tables 4–7).

Discussion

The current study was the first to link retaliatory propensity to reactive aggression in youth with DBD. In addition, it provided the first demonstration of increased PAG responding to provocation in youth with DBD−CU, but not DBD+CU. Critically, this study demonstrated the role of dysfunction of vmPFC and vmPFC-amygdala functional connectivity in regulating retaliatory behavior. Furthermore, these results suggest that disruption in vmPFC-amygdala functional connectivity accounts for the common risk of increased retaliation/reactive aggression in DBD irrespective of CU traits.

In line with our first prediction, retaliatory propensity on our UG task was significantly more positively associated with propensity for reactive relative to proactive aggression. Patients with DBD retaliated more severely during a UG particularly to slightly unfair offers than healthy youth. These data are consistent with other behavioral findings in antisocial adolescents.⁸ In contrast, level of CU traits did not predict level of reactive aggression (though it was associated with proactive aggression), consistent with previous research.³

According to our position,¹⁶ while youth with DBD+CU and DBD−CU both show increased retaliatory behavior, the neuro-biological underpinnings of this increase differs by CU level. Consistent with our second prediction, youth with DBD−CU showed exaggerated basic threat circuitry (PAG/amygdala)³⁵ activation as a function of level of retaliation relative to youth with DBD+CU and healthy youth. These data extend previous work reporting increased amygdala responses to provocation in youth with DBD−CU relative to DBD+CU/healthy youth.^18–20 According to our position, this heightened basic threat circuit sensitivity increases the likelihood of reactive aggression (rather than freezing or flight) in response to a threatening/provocative stimulus.¹⁶ However, it should be noted that neither amygdala nor PAG response predicted general propensity for retaliation in the task or reactive aggression levels more generally. Whether this reflects a type II error or is indicative that this pathology is more related to other associated symptomatology (e.g., anxiety) will be determined in future work.

Our third prediction was that youth with DBD irrespective of CU level are at increased risk for reactive aggression because of dysfunction in vmPFC’s putative role in instrumentally selecting the form of retaliatory behavior as a function of expected rewards/punishments.^5,15,36 It is possible, however, that vmPFC responsiveness on this task might alternatively represent other putative roles for this structure e.g., in directly suppressing the amygdala,³⁷ self-referential processing³⁸ and/or representing social and emotional Structured-Event-Complexes.³⁹ We predicted that patients with DBD would show less reduction in vmPFC activity as a function of punishment level (i.e., less representation of the cost of retaliating).¹⁴ This hypothesis was confirmed for patients with DBD−CU, but not DBD+CU. However, both patients with DBD+CU and DBD−CU showed reduced functional connectivity between right amygdala and vmPFC during high provocation trials relative to healthy youth. This suggests a failure in the appropriate interaction of amygdala and vmPFC in patients with DBD, irrespective of CU level, when responding to high provocation. In line with our fourth prediction, both modulated vmPFC response and vmPFC-amygdala functional connectivity contributed unique variance in the prediction of retaliatory propensity and a mediation analysis supported the critical role of vmPFC in regulating retaliation/reactive aggression. These data suggest that vmPFC-amygdala connectivity is critical for regulating retaliation/reactive aggression describing putative common impairment in youth with DBD+CU and DBD−CU that contributes to retaliatory behavior.

Several caveats should be considered with respect to these data. First, an ADHD comparison group was not included. Mitigating this limitation, a group analysis excluding DBD youth with comorbid ADHD revealed extremely similar results with similar activations for all contrasts. Second, the medications of two youth with DBD could not be withheld for scanning. Again mitigating this limitation, group analyses excluding these participants again identified similar regions for all contrasts. Third, the three-group analyses utilized relatively small samples. Thus it is possible that type II error may account for the failure of youth with DBD+CU to show the hypothesized problems in BOLD response within vmPFC as a function of punishment level (though they did show the predicted reduced vmPFC-amygdala connectivity in response to high provocation). The results of the current study will need to be replicated with larger samples. Fourth, PAG is a small structure that is not expressly defined in AFNI’s probability maps. Therefore, we cannot be certain that the signal detected in PAG is exclusively from PAG; other neighboring structures may also be involved.

In summary, there are three features of the current paper that are particularly important for our understanding of disruptive disorders: (i) The current study was the first to document increased PAG response to provocation in youth with DBD−CU, but not DBD+CU. These data suggest that interventions designed to reduce emotional responsiveness would only be most efficiently applied to youth with DBD−CU; (ii) The current study indicates the critical role of ventromedial frontal cortex in the regulation of retaliatory behavior and that this is dysfunctional (at least its interaction with the amygdala) in patients with disruptive behavior disorders irrespective of level of CU traits. This dysfunction would appear to represent an important treatment target for future interventions; and (iii) level of vmPFC dysfunction on the UG task predicted level of reactive aggression in youth with DBD. This is important, as no fMRI markers of reactive aggression have been previously identified. Moreover, it is important to remember that the dysfunction seen here in patients with DBD is likely seen in patients with other psychiatric disorders who are at risk for mal-adaptively increased levels of reactive aggression; e.g., Posttraumatic Stress Disorder, Borderline Personality Disorder and Disruptive Mood Dysregulation Disorder.⁴⁰ As such, the current study provides a marker task of potential relevance to researchers who are concerned about mal-adaptively increased levels reactive aggression in patients with DBD and patients with these other disorders.

Table 2.

Brain Regions Demonstrating Differential Functional Connectivity in 28 healthy youth, 15 youth with DBD+CU and 15 youth with DBD-CU.

Coordinates of Peak Activation b
Region a	Left/Right	BA	x	y	z	F	p	Voxels
Left Amygdala
Diagnosis-by-Provocation Level
dorsomedial prefrontal gyrus	Left	6	−4.5	4.5	56.5	15.74	<.0001	132
Right Amygdala
Diagnosis-by-Provocation Level
ventromedial prefrontal cortex	Right	32/10	1.5	43.5	5.5	9.34	.0003	22
dorsomedial prefrontal cortex	Left	8	−4.5	28.5	44.5	11.98	<.0001	70
middle temporal cortex	Right	37	52.5	−61.5	5.5	10.20	<.0001	50
superior temporal gyrus	Right	21	58.5	−40.5	−0.5	13.68	<.0001	100
paracentral/cingulate cortex	Right	24/32	25.5	−40.5	50.5	11.36	<.0001	551
parahippocampal gyrus	Right		37.5	−22.5	−12.5	12.94	.0002	40
fusiform gyrus	Right	37	28.5	−46.5	−12.5	12.41	.0008	30
postcentral gyrus	Left	3	−25.5	−31.5	53.5	13.28	<.0001	98
declive	Right		37.5	−70.5	−21.5	12.88	<.0001	89
claustrum	Left	13	−37.5	−22.5	−0.5	12.41	<.0001	88
precentral	Left	6	−37.5	−4.5	53.5	9.19	.0004	73
declive	Left		−49.5	−58.5	−21.5	13.31	<.0001	72
cuneus	Right	19	28.5	−82.5	29.5	8.51	.0006	40
precentral gyrus	Right	4	31.5	−19.5	38.5	8.29	<.0001	30

^aAccording to the Talairach Daemon Atlas (http://www.nitrc.org/projects/tal-daemon/).

^bBased on the Tournoux & Talairach standard brain template, BA= Brodmann’s Area