Aggression is defined as behaviors in which physical force is used with the intent to harm or damage (1). Aggression can be purposeful, unemotional and controlled (i.e., predatory aggression), or manifest itself in reaction to a perceived threat in an emotionally charged situation [i.e., impulsive aggression (IA)] (2). IA manifests itself verbally or physically (3) and is the endpoint of a complex psychopathological cascade starting from hostility. Hostility is a cognitive construct defined as an attitude that involves the dislike and the negative evaluation of others and others' intentions (4). Such attitudinal trait predisposes individuals to respond to external cues with frequent and intense angry feelings and related behavior (4). Therefore, hostility preludes anger, i.e., a multifaceted construct consisting of physiological (general sympathetic arousal), cognitive (irrational beliefs, automatic thoughts, inflammatory imagery), phenomenological (subjective awareness and labeling of angry feelings) and behavioral (facial expressions and verbal/ behavioral anger expression strategies) facets (5–7) Anger in turn serves as a cue for developing physical or verbal aggression (8). IA is present across psychiatric diagnoses (9), but is especially present in bipolar disorder (BD) (10).

BD is a chronic and debilitating illness characterized by cyclic alternations of phases of elated mood, increased energy, reduced need for sleep, increased self-esteem, flight of ideas, distractibility and increased involvement in goal directed activity and, phases of depressed mood, feelings of guilt, motor retardation or restlessness, insomnia, and suicidal ideation (11). The onset of BD is usually in late adolescence/early adulthood, although it can also affect children (12, 13). Juvenile-onset BD, or pediatric bipolar disorder (PBD) has a more severe illness course than adult-onset BD, with higher rates of chronicity, cyclicity, and poorer prognosis (14). IA is noted to be a frequent characteristic in individuals with PBD. Indeed, Hernandez et al. (15), reported IA to be present in more than 50% of 7-year olds with PBD. In young adolescents, these rates (16) almost reached 40%. Rates of IA are higher in PBD compared with ADHD (17) or juvenile depression (16), and have been addressed as one of the most common “red flags” that should alert clinicians for the presence of BD in children and adolescents (13). Papolos et al. (18) included IA among a cluster of symptoms identifying a specific phenotype of PBD characterized by a poor prognosis. Furthermore, our group has shown that IA is associated with impairment in specific cognitive domains, such as affective processing and executive functions (19). Finally, IA is regarded as a common and major public health issue that increases the risk of substance abuse, suicidal behaviors, incarceration and violence in adulthood (20).

Despite the pivotal importance of IA in youth with PBD, its neural correlates have been poorly understood. Specifically, alterations in the cortical sheet, which changes with age (21, 22), in neurological (21, 23) and psychiatric disorders (24, 25), and represents a promising in-vivo biomarker proven useful for diagnosis (26), has not been investigated yet. The extant literature on this topic has mainly focused on healthy youth (27–32) with additional reports in youth with autism spectrum disorders (33) conduct disorder (CD) (30) or attention-deficit and hyperactivity disorder (ADHD) (34). These findings highlighted that, independently from the diagnosis, IA is related to thinning of areas of frontal and parietal lobes. Specifically, cortical thinning was reported in areas belonging to the affective network, i.e., an extensive network involved in mood experience and regulation; and to the control network, i.e., a hub of interrelated networks involved in monitoring performances and control impulses and behavior (35–39). Among all the areas involved, thinning in the anterior cingulate cortex (ACC), posterior cingulate cortex (PCC), orbitofrontal cortex (OFC), supramarginal gyrus (SMG) and precuneus were repeatedly reported. Furthermore, all the aforementioned studies limited the analyses of the neural correlates of IA to the whole construct, and to date, possible differences in cortical thickness and all the facets of IA need to be investigated.

The aim of this study is to evaluate the relationship between cortical thickness, IA, and its facets in youth with PBD. Given the transdiagnostic nature of IA, and its relationship with thinning in areas belonging to the affective network and to the control network, we expected youth with PBD to show inverse correlations between the several facets of IA and the cortical thickness of areas belonging to these networks. More specifically, given the most robust findings regarding thinning of ACC, PCC, OFC, SMG, and the precuneus in aggressive youth, we expected a negative relationship between levels of the different facets of IA and thickness of the aforementioned areas.

This study was approved by The Baylor College of Medicine Institutional Review Board. Study participants were recruited from the child and adolescent outpatient psychiatric clinic at Texas Children's Hospital in Houston between January 2016 and October 2019, whereas HCs were recruited through Craigslist. Parents/legal guardians of youth who met study criteria as determined by a phone screen were approached for participation. Research staff explained the study and study procedures to the parental/legal guardians and children/adolescents. Written informed consent from a parent/legal guardian and written assent from the youth were obtained before any study procedure was initiated.

Sixty-two study participants entered the study. Eight subjects were excluded because they did not fulfill inclusion criteria. Additionally, other 6 subjects were excluded because of motion artifacts. Therefore, the final sample consisted of 46 study participants.

Shapiro-Wilks test showed that variables analyzed have a normal distribution, with the exception of IQ in PBD (W = 0.900; df = 23; p = 0.03). Therefore, Mann–Witney test was performed for investigating between-group differences in IQ. For the other continuous variables, t-tests were performed.

Results might be summarized as follows: (i) youth with PBD showed heightened levels of anger and hostility compared with HC. (ii) youth with PBD showed cortical thinning in prefrontal and parietal cortices. (iii) in youth with PBD, thinning of these cortices was associated with higher levels of anger and lower levels of hostility.

These findings are in line with those reporting heightened anger and hostility (58, 59) and low level of interpersonal violence despite the high levels of incarceration in youth with PBD (60, 61). Additionally, these findings replicate those reporting altered thickness in ACC and PCC (62), OFC (63, 64), inferior parietal cortex (IPC) (65), lingual cortex (LC) and supramarginal gyrus (SMG) (66) in youth with PBD even though cortical thinning has been found in other regions, such as the precentral and superior frontal (67) cortices, or temporal regions (62). Indeed, discrepancies with these findings might be due to the small sample size of our study, or the sample selection of the other studies, which mainly included young adults. Further studies, with larger sample sizes and narrower inclusion criteria are necessary to define alterations in cortical thickness in youth with PBD.

Thinner cortical areas in youth with PBD, i.e., ACC, PCC, OFC, IPC, SMG, LC, are part of two interconnected networks involved in emotion regulation and impulse control. The rostral part of the ACC and the medial and lateral parts of the OFC belong to the affective network (AN) (68–70). The OFC and ACC are densely interconnected with the amygdala and the other subcortical limbic and paralimbic regions through specific feedback and feedforward pathways that serve to convey attention to motivationally relevant stimuli and provide information about the emotional salience or significance of external stimuli in order to rapidly perceive reward contingency (35). Furthermore, both areas participate in the voluntary regulation of emotionally relevant states (36). The other regions belong to a broad network involved in the initiation, maintenance and control of goal directed behavior, and indirectly implicated in impulse control (37, 38). Such “control network” embeds several functions including attention, working memory, response selection, response inhibition, and task switching, that are activated over time in order to prevail against distraction and respond quickly to unpredictable demands that arise during performances (39). This network can be subdivided in two functionally distinct, and partially overlapping networks: the frontoparietal network (FPN) and the cingulo-opercular network (CON). The CON includes the dorsal1 ACC, the PCC, and the LG and contributes to the flexible control of human goal-directed behavior through the stable, across-trial implementation of task sets in downstream sensorimotor processors (37). On the other hand, the FPN includes the IPC, middle cingulate cortex, and supports control initiation and provides flexibility by adjusting control in response to feedback (38).

The CCA demonstrated that the indirect evidence of dysfunction in such networks, i.e., cortical thinning, is correlated with both CDRS and YMRS mean scores. Accordingly, alteration of the AN has been related to manic, depressive, and mixed states, in which depressive and manic phases co-occur (71, 72). Furthermore, the core clinical features of mixed states i.e., irritability, impulsivity and anger, have been related to alterations of both FPN and CON (73, 74). The CCA also show that cortical thinning of areas belonging to the AN, FPN and CON are associated with higher levels of anger, whereas, an opposite pattern was found with hostility. Anger and hostility are considered as two distinct constructs. Anger is seen as an emotional state consisting of physiological, phenomenological and expressive behavioral variables (6, 7) whereas hostility is regarded as more of a cognitive construct (75). Hostility involves cognitive variables of cynism, mistrust and denigration (76), which are mainly driven by a tendency toward viewing others' actions as hostile and purposeful when their intention is unclear, i.e., the hostile attributional bias (8). Activation of the AN, and more specifically, selective hyperactivation of the OFC, has been related to the development of preference for evaluating ambiguous social cues as hostile (77). On the other hand, activation of FPN and OCN has been related to detection of threat (78) and attention bias toward threat (79). Therefore, greater thickness of areas related to AN, FPN, and CON, possibly reflecting greater activity of these networks, might produce greater bias toward negative stimuli and greater alert toward the others, i.e., greater hostile attributional bias. This might turn in greater levels of hostility.

On the other hand, negative correlations between anger and thickness of cortical areas belonging to AN, FPN, and CON might reflect poor activity of these networks as anger outbursts increase. As mentioned before, altered activity of cortical areas belonging to AN are involved in emotion regulation. Poor emotion regulation does not allow for the inhibition of prepotent emotional impulses coming from subcortical areas, such as the amygdala (80). Poor activity in the FPN and CON might facilitate the transition of prepotent emotionally states into action through failure of response inhibition processes., i.e., the capability of withholding or canceling routinized, ongoing, or prepotent responses to enable goal-oriented behaviors (81, 82). Altered emotion regulation and response inhibition might allow emotional negativity to bypass deliberate processing and lead to anger outbursts.

This work has several limitations. First, the small sample size limits the generalizability of the results found. Therefore, larger sample sizes are needed to clearly investigate the relationship between cortical thickness and impulsive aggression in youth with PBD. Second, a narrower age range in youth is needed to investigate the interplay between brain regional developmental trajectories and psychopathology. The broad age range in the present study did not account for brain regional developmental maturation differences. In fact, regional gray matter thickness' plateau varies with age, spanning from 7 to 11 years old, and follows a maturational pattern that could be linear, quadratic or cubic (83). Additional studies with sample sizes including only children or adolescents are needed to investigate whether the present findings are applicable to either children and adolescents. Third, the effect of possible confounding variables might have interfered with the results found. Such variables have been proved to modify brain and behavior and include, but are not limited to specific medications (84–86), childhood trauma (87, 88) or predominant polarity (89). Specific work on the effect of these variables on the neural bases of impulsive aggression are warranted to clarify their impact.

This preliminary work highlights that altered thickness in areas belonging to the AN, PFN and CON are related to the different facets of impulsive aggression. Specifically, results suggest that hyperactivation of the AN, FPN, and CON might result in greater levels of hostile attributional bias, thus increasing levels of hostility. On the other hand, hypoactivation of the aforementioned structures are related to affective dysregulation and behavioral disinhibition, which can result in anger outbursts. Future research and larger samples, using different MRI techniques, are needed to clarify dysfunctional patterns underlying impulsive aggression and its multiple facets in youth with PBD. Novel treatments specifically restoring balance in the activity of such networks are warranted to prevent the development of hostile behaviors and anger outbursts in children and adolescents with PBD.

The raw data supporting the conclusions of this article will be made available by the authors upon request, without undue reservation.

The studies involving human participants were reviewed and approved by Baylor College of Medicine Institutional Review Board. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

KS, GS, and AS designed the study. AS, SK, and JS acquired data. CV and MDN analyzed the data, together with KS and SK. AS, AR, JCS, and GS drafted the manuscript. KS, AR, CV, MDN, and JCS revised the manuscript. All the authors approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.