Introduction
Callous-unemotional (CU) traits (e.g., a lack of remorse, decreased empathy and callousness) are associated with a more severe and chronic trajectory of aggressive antisocial behavior [
1,
2]. These traits may serve as a precursor to psychopathy [
3] and persistent antisocial behavioral problems in adulthood [
4,
5]. Additionally, treatment effects may be lower in adolescents with antisocial behavior and CU traits compared to youth with antisocial behavior, but without CU traits, although this may also be the result of poorer premorbid functioning [
6]. The importance of CU traits is further highlighted by the addition of a specifier to the diagnosis of conduct disorder in the DSM-5 [
7].
The etiopathogenesis of CU traits, however, is still largely unknown. There is a growing body of literature stating that adolescents with CU traits represent a heterogeneous sample, with multiple developmental pathways. First, there is the primary variant, high in CU traits and low in anxiety, which is proposed to be a temperamental or genetically based deficit in emotion processing. Second, there is the secondary variant, high in CU traits and high in anxiety, which etiologically could be the result of childhood abuse or trauma exposure [
3]. Low anxiety levels in primary CU variants appear to arise from a fearless temperament and lack of emotional response to caregivers when committing transgressions [
8,
9]. Adolescents high in CU traits and low in anxiety also show reduced fear potentiated startle [
10], decreased amygdala activity, as well as skin conductance in response to negative faces [
11]. On the other hand, adolescents high in CU traits and high in anxiety show hyperarousal and oversensitivity to negative affect, which is a typical consequence of early childhood maltreatment and traumatization [
12]. Emotional numbing and detachment may serve as protective coping strategies to adverse childhood experiences [
12], reducing stress levels in the short run, but stimulating aggressive and antisocial behavior in the long run, by suppressing empathy and moral socialization [
13].
The proportion of primary and secondary variants among adolescents varies widely, with higher rates of primary variants in justice-involved samples compared to clinical/high-risk community samples [
3]. Yet, some studies failed to distinguish primary and secondary variants [
14,
15]. Differences between studies might be attributed to different statistical procedures used to identify primary and secondary variants, most commonly person-centered clustering techniques, but also mean-split [
16] or moderated regression analyses [
17,
18]. Studies have also been heterogenic regarding the use of CU traits or the broad dimensional psychopathy concept, and in measurement instruments for anxiety [
3].
Early identification of CU-traits is crucial to promote prevention strategies and to provide early interventions to youth-at-risk and their families. Further classification of adolescents with primary vs. secondary CU variants may also affect prognosis and treatment outcome. Since the secondary CU variant was repeatedly demonstrated to be more related to a history of child abuse and trauma than the primary variant [
16,
19,
20], adolescents with secondary variants might benefit from trauma therapeutic intervention, whereas adolescents with primary variants might not. With regard to preventive strategies, we need detailed information on prevalence rates of adolescents with primary vs. secondary traits in (high-risk) community samples, as well as on the burden of disease and clinical correlates in both groups. Current evidence shows that adolescents with secondary CU variants display more overt antisocial (i.e., aggressive) behavior (e.g., physical assault) compared to adolescents with the primary variant [
4,
19,
20]. There is a lack of population-based studies on CU variants and covert antisocial behavior (i.e., rule breaking behavior such as lying, cheating, arson, vandalism, and theft); in detained youth increased rates were found in adolescents with the secondary CU variant compared to the primary variant [
20]. The importance of distinguishing overt and covert antisocial behavior is further demonstrated by a long-term follow-up study into adulthood, showing non-violent crime convictions are a common consequence to covert antisocial behavior in mid-adolescence, signaling a covert developmental pathway [
21].
Another gap in the literature on primary/secondary CU variants concerns the association with risk of first episode psychosis. We found one study in which adolescents from a community sample with secondary CU traits exhibited more symptoms of psychoticism compared to the primary CU subgroup [
19]. Adolescence is a vulnerable transition period in which individuals at risk can develop psychotic symptoms, early-onset psychosis or schizophrenia. Different phenotypes have been proposed regarding the relationship between psychotic disorders and antisocial/aggressive behavior: antisocial behavior may precede and accompany psychosis onset, aggressive behavior may start with psychotic illness progression, and/or an aggressive event may occur after many years of illness [
22]. The role of CU traits in either trajectory is not clear, thus investigating associations with different CU variants at a developmental stage where transition into first episode psychosis is at stake is of particular importance.
Furthermore, to our knowledge, only one study investigated substance use and the associations with CU variants in a community sample: in a single gender study Goulter et al., found that girls with secondary CU traits showed significantly higher levels of substance use than a healthy control group without CU-traits, but differences between girls with secondary and girls with primary CU-traits were not statistically significant [
23].
Correlates of primary or secondary CU traits and addictive behaviors such as internet gaming disorder have not been investigated either. In DSM-5 internet gaming disorder (IGD) has been included in section III, highlighting the need for further clarification as a diagnostic construct [
7]. Internet gaming disorder is defined as a continuous and repeated involvement with video games, often leading to significant daily/educational/social disruptions. Nine diagnostic criteria have been proposed including withdrawal symptoms when not playing, tolerance, continuation of playing despite problems, and playing as an escape of adverse moods. Prevalence rates in adolescents vary between studies (1–6%) [
24,
25]. IGD has been associated with a higher degree of attention/hyperactivity symptoms, emotional distress, and antisocial behavior in adolescents [
26], though a lack of association with psychopathology has also been found [
27]. Clarification of the potential association of IGD and CU traits could mark protective or threatening comorbidity in adolescents at risk.
Other manifestations of antisocial behavior are bullying perpetration and victimization. In a recent meta-analysis, it was found that CU traits are positively associated with bullying perpetration, and also with bullying victimization [
28]. None of the included studies investigated primary and secondary variants. Disentangling associations with primary and secondary CU variants may better indicate targets for tailored interventions against bullying.
In the present study, we examined callous-unemotional traits with or without anxiety features in a high-risk community sample of adolescents, who were oversampled on the basis of self-reported emotional and behavioral problems. Our first aim was to cluster different subgroups based on various types of overt and covert antisocial behavior using latent profile analysis (LPA). We hypothesized to find a class of adolescents with antisocial behavior which would represent a primary CU variant (i.e., high level of CU traits and low level of anxious symptoms), a secondary CU variant (i.e., high level of CU traits and high level of anxious symptoms), an anxious variant (low levels of CU traits and high levels anxious symptoms), and a reference group with low levels of both CU traits and anxious symptoms. We expected the highest levels of overt and covert antisocial behavior in the secondary CU variant compared to the primary variant and the non-CU groups. The secondary aim of this study was to address existing gaps in knowledge on clinical correlates, therefore we explored whether substance use, internet gaming addiction, psychotic experiences, and bullying perpetration/victimization varied among these different CU variants.
Discussion
In the present study we investigated the presence of CU traits in a Dutch high-risk community sample of male and female adolescents and identified different subgroups based on rule-breaking and aggressive behavior and gender. Using latent profile analyses, we found three classes of adolescents that significantly differed in antisocial behavior: a high CU-low anxiety group, an intermediate CU-high anxiety group, and a reference group (low CU-low anxiety).
Contrary to our hypothesis and earlier studies in high-risk and clinical samples [
16,
23,
54], a three-class- but not a four-class model- was found. This could either be the result of studying rare features like CU traits in a community sample, or of the use of two indicators (i.e., CU traits, anxiety) and one covariate (i.e., gender) in the LPA procedure. Of note is that some studies in high-risk community samples finding a fourth high anxious-low CU traits variant used a different statistical method, median-split cut-off [
16] or used a female-only sample [
23]. Three classes were also found in a mixed clinic-referred sample using model-based cluster analysis [
20]. Our prevalence rates were comparable to those earlier described in a large community sample [
4], who found figures of 7.8% for a primary CU variant, 8.9% for an anxious variant, and 2.8% for a secondary variant. Other studies dealing with primary and secondary variants in high-risk or clinical populations [
17,
23,
55] used selected samples of adolescents with problem behavior or developmental adversity. Our intermediate CU-high anxiety group is likely a mixture of adolescents corresponding to the secondary CU variant, an emotionally unstable subgroup of adolescents with primarily reactive aggressive behavior, and an adolescence-limited delinquency subgroup. In sum, our results significantly add to the knowledge of detecting at-risk individuals in the community, e.g., school settings, by scrutinizing CU traits, anxiety features, and problem behavior as warning signs that call for intervention.
The intermediate CU-high anxious group yielded highest scores on aggressive behavior, whereas the scores on other types of antisocial behavior scores were comparable to the primary variant group; for property offenses the primary CU group scored highest. This combined elevated risk of both overt and covert forms in both groups is suggestive of further risk for criminal offenses, as combined overt and covert antisocial behavior is associated with high stable antisocial behavior at age 18 [
56]. Although the primary CU group is often considered at prime risk for future offenses [
5,
57], there is increasing evidence from longitudinal studies that antisocial behavior with comorbid anxiety symptoms confers equal risk of future mental health problems and offenses [
58,
59]. Furthermore, extending previous reports [
59,
60], the intermediate CU-high anxious group might experience more peer rejection, as witnessed by bullying victimization, and reactive aggression, while the primary CU group might be more prone to popularity striving and proactive aggression.
Our results further expand knowledge by showing an association between CU traits and high anxiety traits in adolescents with antisocial behavior on the one hand and psychotic experiences on the other. The intermediate CU-high anxious group reached particularly high scores on psychotic experiences. Manifestation of first-episode psychosis is in about one third of patients accompanied by aggressive behavior [
61]. Common distal risk factors for disruptive behavior disorders and schizophrenia such as obstetric complications, neurodevelopmental delay and learning disabilities, together with genetic vulnerabilities, may increase risk for both behaviors. Alternatively, psychotic experiences at this age may serve as markers of a wide range of internalizing and externalizing psychiatric symptoms [
62]. Though speculative, adolescents in the community with a combination of anxious symptoms, antisocial behavior and psychotic experiences could be at increased risk for future offenses or forensically relevant psychopathology. A study in convicted adults shows that a subgroup with childhood antisocial behavior and increased CU traits and anxiety are at higher odds of a diagnosis of borderline personality disorder, which can be marked by transient paranoid ideation [
58].
The primary CU group scored higher than other groups on IGD. Despite contradictory results in earlier studies, a recent meta-analysis suggests IGD has been associated with several externalizing disorders, including aggressive and rule breaking behavior [
63]. High CU traits can operate as both an antecedent and outcome of IGD, in that adolescents with high CU traits tend to have difficulties setting up social interactions with peers, while excessive gaming leads to avoiding real world interactions in which empathy can possibly be stimulated. Interestingly, preferred game types are differentially associated with personality features. Low empathy has been particularly found in role playing games (e.g., ‘Roblox’), whereas real world competences like leadership and self-understanding may be enhanced in Massive Multiplayer Online Role Playing Games (MMORPG) (e.g., ‘World of Warcraft’) [
64]. Whether excessively playing videogames by adolescents prone to CU-traits increases the risk, severity or stability of antisocial behavior warrants further investigation.
The intermediate CU/high anxious group scored significantly higher than the other groups on bullying victimization. This finding fits with numerous studies finding exposure to abuse and trauma as a fundamental aspect of secondary variant etiology and posing individuals at risk of repeated victimization [
3]. Besides, as our intermediate CU-high anxious group is no classic ‘secondary’ CU variant, adolescents reporting higher anxious symptoms without high CU traits are more likely to suffer from bullying victimization, and associated loneliness and low self-esteem [
65].
The findings in this study need to be viewed in light of some limitations. First, although we included several instruments by different informants to assess antisocial behavior, we cannot exclude the risk of some reporter bias. Adolescents were the only informants on CU traits, violent and property offending, psychotic experiences and videogaming. Earlier studies revealed that adolescents with primary variant CU traits are likely to underreport severe types of antisocial behavior [
20]. Second, due to selection bias, we may have tapped on a subgroup of adolescents high in CU traits but without high levels of antisocial behavior [
66], as responders included in the analyses were more likely to have average intelligence, SES and were less likely to belong to ethnic minorities compared to non-responders. As a result, differences in level of CU traits and severity of antisocial behavior between the high CU and the intermediate CU-high anxious group may have been obscured. Still, our results generally fit in those of previous studies comparing aggressive behavior between primary and secondary variants in community and clinical settings [
19,
20]. Third, due to the cross-sectional design, our study cannot provide direct information on the developing trajectory of (innate) CU-traits, adverse childhood experiences, anxious symptoms, other clinical markers, and antisocial behaviors. Moreover, pinpointing proximal or distal precursors over time is difficult with assessments scoping on symptoms in the previous 6 months. Caution in interpreting the results is also warranted as CU traits were measured approximately 9 months after all other assessments. Although most adolescents at mid-adolescence show stable patterns of CU traits, we cannot rule out that a small subsample either had a decreasing or increasing pattern during this time window.
In conclusion, our study provides further evidence for classification of CU variants, a high CU variant and an intermediate CU-high anxious variant, in adolescents who have not been in contact with the judicial system. Future prospective studies should point out whether and to what extent adolescents with CU traits with and without anxiety develop criminal careers and adult psychiatric disorders. Intermediate CU traits combined with high anxiety (even in the presence of high levels of aggressive behavior) may be as predictive for future delinquent behavior as high CU traits and low anxious traits. Early recognition in schools of youth with combined traumatization/anxiety symptoms and behavioral problems can be done by routine screening, though establishing cut-off levels remains challenging. These adolescents may benefit from thorough assessment, including internet gaming addiction and psychotic experiences. Moreover, the association of CU traits and internet gaming disorder needs further clarification in terms of loss of control and violent aspects of game playing. As elevated CU traits are related to reduced facial reactions of sadness and disgust to violent films [
67], the question is raised to what extent violent game playing in mid-adolescence fuels the risk of later violent aggressive acts, particularly in children with primary CU-trait variants. In sum, early identification of youth with CU/anxious traits may better guide proposals for intervention.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.