Introduction
Children and adolescents with conduct problems (CP) constitute a heterogeneous group, not only in types of CP they exhibit (Lindhiem et al.
2015), but also in their risk for future antisocial outcomes (e.g., Odgers et al.
2008). Past research suggests that callous-unemotional (CU) traits help to identify a subgroup of children with CP who exhibit a more severe and stable pattern of CP compared to youth with CP low on CU traits (Frick et al.
2014). Reflecting this body of evidence, CU traits have increasingly been included in theoretical models and empirical studies on CP, and are expected to influence clinical work with children and adolescents, especially since classification systems have added (
DSM-5), or may add (
ICD-11) a CU-based specifier for the diagnosis of conduct disorder (APA
2013; Salekin
2016a,
b,
2017). Notwithstanding the relevance of studying CU traits in relation to CP, and as detailed elsewhere in this Special Issue (see Colins, Andershed, Salekin, & Fanti
2018) it remains unclear if CU traits is the best predictor of future and stable antisocial outcomes, or if a greater representation of psychopathic traits is needed to identify the adolescents who are at the highest risk.
A recently published study among 1867 3- to 5-years old Swedish boys and girls used teacher ratings of CP and three psychopathic traits dimensions (i.e., CU, interpersonal and behavioral/lifestyle) to assign children to six mutually exclusive groups (Frogner, Gibson, Andershed, & Andershed
2016). These groups were: (1) low on CP and all three psychopathic traits dimensions (
Control); (2) high on CP and low on all three psychopathy dimensions (
CP Only); (3) low on CP and high on the CU traits dimension only (
Callous-Unemotional Only); (4) low on CP and high on all three psychopathic traits dimensions (
Psychopathic Personality Only); (5) high on CP and on the CU traits dimension only (
Callous-Unemotional + CP); and (6) high on CP and all three psychopathic traits dimension (
Psychopathic Personality + CP). Crucially,
Psychopathic Personality + CP boys were at a greater risk for future and stable CP than the other groups. Overall, this finding was replicated among girls with the notable exception that
Psychopathic Personality + CP and
Callous-Unemotional + CP girls were equally at risk for stable CP. Colins and colleagues (
2018; This special issue) tested if the findings from Frogner et al. (
2016) could be replicated among 690 7- to 12-year old Cypriot boys and girls whilst using the same analytical approach but using other informants (i.e., parents) and other psychopathy measures. Results showed that
Psychopathic Personality + CP children by far showed the most robust and highest risk for future and stable CP, whereas
CP Only children were often equally, and sometimes even at a higher risk than
Callous-Unemotional + CP children. Taken together, there is some evidence to suggest that using CU traits only is less sufficient than using CU traits in combination with the other psychopathic traits dimensions for identifying CP youth who are at the highest risk for future and stable CP. Prior work (Colins et al.
2018) also revealed significant prospective relations between
Psychopathic Personality Only – and to a lesser extent also between
CU Only – on the one hand and antisocial behavior on the other hand, suggesting that psychopathic traits that do not co-occur with baseline CP have at least some prognostic utility.
This Study
The present study was designed to replicate and extend these aforementioned findings in various ways. First, this study will rely on self-report measures instead of teachers (Frogner et al.
2016) or parents (Colins et al.,
2018, this Special Issue). Self-report measures constitute a major advancement in the assessment of psychopathic traits (e.g., Colins and Andershed
2016; Vitacco et al.
2014), and, thus, are also important to put CP subtyping approaches to the test (e.g., Colins and Andershed
2015; Kimonis et al.
2015). Second, antisocial behaviors are quite heterogeneous in kind (e.g., aggression, rule-breaking) and severity (e.g., fighting, lying, shoplifting). As such, it is relevant to know if CP subtyping approaches also help to identify youth at risk for one of the most severe forms of antisocial behavior, that is, aggression against other people. In addition, substance use has been a relevant external variable to evaluate in subtyping models (e.g., Frick et al.
2014; Wymbs et al.
2012) and can be considered an important outcome when scrutinizing the prognostic usefulness of psychopathy scores (Colins et al.
2015). Therefore, this study will also include antisocial outcomes other than CP, including aggression and substance use.
Hypotheses
Mirroring the expectation that children with CP who manifest psychopathic traits constitute a severe CP subgroup (e.g., Frick et al.
2014; APA
2013), it was first hypothesized that
Psychopathic Personality + CP and
Callous-Unemotional + CP adolescents will show the highest CP during baseline than adolescents in the other groups. Crucially, it was also expected that the
Psychopathic Personality + CP adolescents will present the highest CP during baseline compared to
Callous-Unemotional + CP adolescents. Second, it is expected that
Psychopathic Personality + CP and
Callous-Unemotional + CP adolescents will be at a higher risk for future and stable CP, aggression, and substance use, than their counterparts in the other groups.
Psychopathic Personality + CP adolescents are nevertheless expected to be at a higher risk than
Callous-Unemotional + CP adolescents. Finally, the present study also explored if adolescents in the
Callous-Unemotional Only and
Psychopathic Personality Only groups are at risk for developing the antisocial outcomes under consideration.
Discussion
The main aim of the present study was to compare CU- and psychopathy-based subtyping approaches in their ability to predict future and stable forms of various antisocial behaviors. The most robust and strongest prospective relation with future and stable antisocial outcomes occurred when the combination of high levels of interpersonal (i.e., Grandiosity), affective (i.e., CU traits), and behavioral/lifestyle (i.e., Impulsivity) traits co-occurred with high levels of CP at baseline (Psychopathic Personality + CP). This main finding and its consistency with prior work suggests that CU traits-based approaches for subtyping children and adolescents with CP are not as efficient as subtyping approaches that use the multidimensional youth psychopathy construct for predicting future and stable antisocial behaviors. The second main finding is that in adolescents without CP, being high on all three psychopathic traits dimensions (Psychopathic Personality Only) often was positively related to future and stable antisocial outcomes, whereas being high on CU traits only (Callous-Unemotional Only) was not. Thus, even in the absence of baseline CP, being high on the three psychopathic traits dimensions is a better predictor of future and stable antisocial outcomes than being high on CU traits only.
Following prior work (e.g., Fanti
2013), adolescents could be assigned to
CP Only and
Callous-Unemotional + CP groups. Importantly, a substantial number of adolescents with CP was also assigned to the
Psychopathic Personality + CP group providing evidence for additional heterogeneity in CP. Comparing these three CP groups in terms of baseline levels of CP shows that – as hypothesized –
Psychopathic Personality + CP youth were higher in baseline CP than the other two CP groups, though it must be noted that the
Psychopathic Personality + CP and the
Callous-Unemotional + CP Only groups were not significantly different in baseline CP after correcting for multiple group comparisons. Of note, only a small number of adolescents with CP were assigned to the
Callous-Unemotional + CP group (
n = 14), whereas 61 adolescents with CP could be assigned to the
Psychopathic Personality + CP group. This finding supports the view that being high on CU traits and being high on all three psychopathy dimensions identifies largely overlapping groups of children with CP (Frick
2009), though it is important to note that
Callous-Unemotional + CP group assignment was not equally predictive as
Psychopathic Personality + CP group assignment. To illuminate how robust this overlap is, much more work is needed, especially since available evidence on this overlap is mixed (e.g., Colins et al
2018; Christian et al.
1997). Unfortunately, we were not able to use more stringent cut-offs (e.g., > 1.00 SD) to assign the participants to mutually exclusive groups due to sample size restrictions. This limitation may explain why the percentage (55%) of adolescents with CP who were assigned to the
Psychopathic Personality +
CP group is too high in light of evidence that the estimated prevalence of Psychopathy Checklist-Revised defined psychopathy in samples of antisocial adults is 15–25% (Drislane and Patrick
2013). As such, future endeavors to find the best way of identifying a relatively small subgroup of CP youth who most strongly and consistently display the features associated with adult psychopathy should be encouraged.
CU traits are considered important for identifying a subgroup of children and adolescents with CP who are at higher risk for future and stable antisocial behaviors (e.g., Frick et al.
2014). Yet, in the present study,
Callous-Unemotional + CP adolescents were only at risk for future CP at two out of three follow-up assessments, whereas
CP Only adolescents were at risk for future CP, future and stable aggression,
and future substance use.
Psychopathic Personality + CP adolescents were consistently at high risk for all outcomes under consideration at each follow-up assessment, and to a much higher extent than adolescents in any other group, including the
Callous-Unemotional + CP group. We see at least three potential explanations for this finding. First, impulsivity is a well-documented predictor of various negative outcomes (e.g., Caspi et al.
1996) and mounting evidence shows that interpersonal psychopathic traits are uniquely related to bullying, delinquency, and aggression, sometimes even stronger than CU traits (e.g., Lau and Marsee
2013; Stellwagen and Kerig
2013; Colins
2016). As such, the difference in predictive ability between the two groups may be explained by the higher baseline levels of Grandiosity and Impulsivity seen in the
Psychopathic Personality + CP group. Second, various studies on the development of antisocial behavior have documented that the higher the number of risk factors, the higher the risk for antisocial behavior (e.g., Stouthamer-Loeber et al.
2002). Thus, the
Psychopathic Personality + CP (versus
Callous-Unemotional + CP) group may be at a higher risk for antisocial behavior simply because they display more co-occurring risk factors. Third, it may be, that it is something in the specific combination of the three psychopathic traits dimensions that put youth at a higher risk for negative outcomes. In support for this notion, several studies have found a three-way-interaction effect between the three psychopathic traits dimensions in relation to concurrent and future antisocial behavior in youth (e.g., Colins et al.
2014; Fanti et al.
in press; Orue and Andershed
2015).
This study also provides information about the prognostic usefulness of psychopathic traits in youth
without CP at baseline (i.e., the
Callous-Unemotional Only and
Psychopathic Personality Only groups). First, CU traits that occur in the absence of CP or other forms of antisocial behavior are considered to bear clinical significance, for example because
Callous-Unemotional Only individuals may be at risk for later antisocial behavior (Viding and McCrory
2012). Using the cut-offs of the present study, we do find a rather large group of youths with CU traits without high levels of CP (
n = 90). The numbers will of course vary with cut-off, but research using other cut-offs and other ways to identify groups, have also found a rather large group of CU only youths (e.g., Fanti
2013). An important question is though whether different cut-offs will yield differences in predictive power. Recent studies testing various cut-offs indicates that different cut-offs does not seem to affect the predictive power (Frogner et al.
2016; Colins et al.
2018; Frogner, Andershed, & Andershed
in press). The present study, and these other recent studies (Frogner et al.
2016; Colins et al.
2018; Frogner et al.
in press) could not support the notion that CU traits only (i.e., without concurrent CP) may be associated with a high risk for future antisocial behavior, neither when predicting future and stable CP, nor when predicting future and stable aggression and substance use. In fact,
Callous-Unemotional Only youth were significantly less likely to display future and stable antisocial outcomes, agreeing with prior work (Fanti
2013).
Our findings may also be informative when debating the role of the antisocial dimension (Skeem et al.
2011). Specifically, participants in the present study who were assigned to a psychopathic personality group were at a higher risk for future and stable antisocial behavior when they had high baseline levels of CP (
Psychopathic Personality + CP) as compared to when they did not (
Psychopathic Personality Only). This finding dovetails well with evidence showing that adult psychopathy is less predictive of future antisocial behaviors when the antisocial dimension is excluded from the total psychopathy score (e.g., Vitacco et al.
2005). Importantly, our study also suggest that even without introducing a prognostic tautology (in our case CP), being high on all three psychopathic traits dimension helps to identify school-attending adolescents at risk for future CP and other forms of antisocial behavior.
The strengths of the present study include the longitudinal design (3-year follow-up), the use of various negative outcomes (Salekin
2008), and the reliance on measures (APSD and ICU) that have been extensively used in prior work on CU traits and youth psychopathy. Our findings however must be interpreted in the context of several limitations. First, when using higher cut-offs such as .75
SD and 1.00
SD above the mean as cut-offs, this resulted in groups with too few participants to directly compare the groups of interest with each other. Prior work however, as mentioned previously, shows that the pattern of findings remains substantially similar when using more stringent cut-offs (Colins et al.
2018; Frogner et al.
2016). Nevertheless, to ascertain that our findings can be generalized when using more stringent cut-off scores that are more likely reflective of severe baseline levels of CP and psychopathic traits, studies with larger sample sizes are needed, preferably among clinic-referred and criminal justice-involved individuals. Researchers who already have large data sets available can start to address this issue. Second, our analytical approach (categorizing continuous variables) was based on arbitrary cut-offs and may have resulted in loss of statistical power and increased probability of committing type-I errors (MacCallum et al.
2002). We acknowledge these arguments but also note that despite this assumed reduction in terms of statistical power, we nevertheless found prospective relations between group-membership categories and future and stable antisocial outcomes. Both continuous and categorical approaches are useful and necessary (Lilienfeld
2014), and we encourage researchers to come up with alternatives and better strategies to compare the CU traits only versus multidimensional psychopathy approaches. Third, the present study could only rely on self-report information, and future research that includes various informants will help to overcome the possibility that our prospective relations are largely explained by shared method variance. Fourth, the present study could not specifically test the questions related to
why there were differences between the various groupings in the prediction of antisocial behavior. A psychopathy based theoretical hypothesis here would be that the
Psychopathic Personality + CP group is worse off because they exhibit multiple and interacting problematic traits and behaviors, as compared to the other groups studied. Some may however question this and propose that it is merely the higher levels of CU traits in the
Psychopathic Personality + CP group that explain that groups´ higher risk for future antisocial behavior. These various views could not be tested with the analytical approach used in the present study. Finally, our study did not test the stability of the group assignments and their relation to various antisocial outcomes, and we acknowledge this as a limitation that should be addressed in future research. Filling this void will be important, for example, to learn if youth in the
Psychopathic Personality + CP group are at a much greater risk to be identified as an adult with psychopathy than their
Callous-Unemotional + CP counterparts, and to see if youth in the
Callous-Unemotional Only and
Psychopathic Personality Only manage to stay under the radar of law enforcement agencies.
The findings of the present study suggest that the entire psychopathy construct outperforms the CU traits alone model in identifying CP youths who are at risk for severe and stable antisocial behavior. These findings together with the fact that psychopathy commonly is defined as a multidimensional construct (e.g., Cooke and Michie
2001; Frick et al.
2000) and often seen as an important risk factor for future antisocial behavior (e.g., DeLisi
2017; Hare
2016) lend support for the notion that researchers may need to consider focusing on the broader psychopathy construct.
If replicated, the present study provides important practical and theoretical information. The findings indicate that the three factor model of psychopathy may be a good definition of psychopathy among youths, since it predicts stable forms of aggression and substance use, albeit not conduct problems. The present study would also indicate that the four factor model of psychopathy (i.e., including concurrent antisocial behavior as a fourth factor) may be a good definition of psychopathy among youths, since this by far was the best predictor of all stable forms of antisocial behaviors in the present study.
Furthermore, in terms of implications for diagnostics and more specifically the Conduct Disorder diagnosis, the present study shows the need for subtyping youths with conduct problems/Conduct Disorder because subgroups with conduct problems (especially the subgroup with concurrent psychopathic personality), show a different level of risk for future antisocial behavior. Youths with conduct problems only (i.e., without concurrent psychopathic traits) showed lower risk for antisocial behavior than the group of youths with conduct problems and concurrent psychopathic personality. In the most recent version of the DSM, only CU traits are used as a specifier for conduct disorder (APA
2013). The current study showed that youth in the
Callous-Unemotional + CP group were at an increased risk for future CP, though not as much as youth in the
Psychopathic Personality + CP group. Crucially,
CU + CP youth were not at an increased risk for future and stable aggression and substance use whereas their
Psychopathic Personality + CP and even their
CP Only counterparts were. Further research on the topic will inform future revisions of DSM-5 and ICD-11 if interpersonal and behavioral/lifestyle traits need to be included as additional specifiers for identifying children and adolescents with a psychopathic personality. Being able to differentiate between
CU + CP and
Psychopathic Personality + CP youth is not only relevant to avoid that youth who merely display CU traits are misclassified with the stigmatizing label “psychopathic personality’, but it also may help clinicians to identify youth who are at greatest need of intensive treatment due to their chronic and severe engagement in antisocial behavior.