Child Neuropsychological Functioning and Interpersonal Callousness as Predictors of Externalising Behaviour in Early Adolescence: A Prospective Population-based Study

One factor that has received substantial attention in relation to externalising problems is neuropsychological functioning (Frick et al., 2014; Rohlf et al., 2018; Thompson et al., 2019). neuropsychological functioning includes a diverse set of abilities including language, learning and memory, social cognition, perceptual-motor function, and executive functions (Séguin & Zelazo, 2005; Thomson & Centifanti, 2018). Several independent studies have reported an association between poorer executive functioning – defined as higher-order cognitive abilities, such as attention, planning, abstract reasoning, inhibitory control and working memory (Jurado & Roselli, 2007; Morgan & Lilienfeld, 2000) – and various forms of externalising behaviour, such as delinquency, criminality and physical aggression (Morgan & Lilienfeld, 2000; Ogilvie et al., 2011). However, potential associations with other neurocognitive domains remain unclear. For example, a recent meta-analysis found that low language ability predicts the later development of externalising behaviour in different samples of school-aged children that differed in risk status and age (Chow et al., 2018). In contrast, a large population-based study including children from the general population found no associations between externalising behaviour at age 6 years (Blanken et al., 2017) and language abilities at age 8 years, although it did confirm an association with poor attention and EF. Further, very few studies have examined both global neuropsychological functioning and different subsets, so that it is unclear whether previously identified associations are domain specific or generalized. Indeed, a recent large longitudinal study of neurodevelopment (The Adolescent Brain Cognitive Development Study) in children from the general population of age 8 and 9 years, found that externalising behaviour was predicted by lower levels of general cognitive ability, not by EF specifically (Thompson et al., 2019).

In addition to low neuropsychological functioning, callous-unemotional traits are also found to be an important predictor of later externalising behaviour (Frick et al., 2005). Callous traits involve characteristics such as being mean to others and not feeling guilty about one's own actions, which are elements of a more extensive set of callous-unemotional traits (CU-traits) and psychopathic traits (Frick, 2004). The DSM-V, which refers to CU-traits as 'limited prosocial emotions', has included this behaviour as a specifier for conduct problems (APA, 2013). CU-traits involve, next to callous traits, shallow emotions, and disregard for the feelings of others (Frick, 2009). Numerous different measures, varying in their inclusion of aforementioned specific behaviours, have been used to study callous-unemotional traits. This needs to be considered when interpreting the existing literature.

Previous research has shown that in a sample of children from the general population, CU-traits at age 3 years predict externalising problems at later ages, namely age 6 years and 10 years, over and above earlier externalizing problems at age 3 years (Song et al., 2016). Additionally, a systematic review including eight different samples from different populations (e.g. general population, clinically referred, detained) and from different age levels (ranging between 5 and 18 years of age), that children with conduct problems and CU-traits are more at risk for severe externalising behaviour later in life than children with CD without CU-traits (Colins et al., 2020).

When looking at the relation between neuropsychological functioning and CU-traits, it has been reported that high CU-traits and high EF predicted greater violent behaviour in a high-risk sample of adolescents (Baskin-Sommers et al., 2015). A more recent population-based study that looked at both intelligence and executive functioning, found CU-traits to be a moderator for the association between sustained attention specifically, and antisocial behaviour in a sample of adolescents (mean age 14 years) from the general population (Dotterer et al., 2021). At high levels of CU-traits, antisocial behaviour in children was associated with better performance on a sustained attention task, whereas it was associated with worse performance with low levels of CU-traits. In contrast to these results, Waller et al. (2017) found that CU-traits at age 3 predict aggression in late childhood and that low levels of EF modulated this effect. These results imply that CU-traits could be a possible moderator of the association between neuropsychological functioning and externalising behaviour, although the findings are mixed.

Mixed evidence about the role of neuropsychological functioning and CU traits in externalising problems could be due to substantial differences in study design (cross-sectional vs longitudinal), timing of variables assessed (childhood vs adolescence) and samples used (e.g. community-based vs offender and high-risk clinical samples). Prospective studies using well-characterized, longitudinal data across development are needed to better understand the associations between EF, CU-traits and externalising behaviour in children. Furthermore, the results from studies that included offenders or high-risk samples may not generalise to the general population. The latter implies that interventions based on the data of offender samples may not be as effective in non-offending children.

Sex could also be important in understanding the association between neurocognitive functioning and externalising behaviour due to sex differences in the occurrence and manifestation of externalising behaviours. For example, it is well-established in the literature that boys show more externalising problems than girls (Arnett et al., 2014; Mayes et al., 2020), and score higher on CU- traits than girls (Essau et al., 2006; Pihet et al., 2015). Further, some studies have shown that sex moderates the association between neurocognitive functioning and externalising behaviours. For example, one study found that the negative association between EF (inhibition) and (proactive) aggression was stronger and only significant for girls (Granvald & Marciszko, 2015). However, this moderating effect was not found for reactive aggression or other forms of aggression in relation to poor EF (McQuade et al., 2013; White et al., 2013b). Thus, it remains unclear whether sex moderates the effect of neuropsychological functioning on externalising behaviour in children.

In summary, although previous research has examined the relation between neuropsychological functioning and externalising behaviour later in life, key gaps in the literature remain. First, previous studies looking into the role of global neuropsychological functioning vs subdomains of neuropsychological functioning are lacking. We want to look at the associations between neuropsychological functioning more comprehensively to disentangle general vs specific associations between neuropsychological functioning and externalising behaviour. Secondly, the moderating role of CU-traits, and especially interpersonal callousness in the association between neuropsychological functioning and externalising behaviour remains unclear. Third, studies have been mainly being carried out in boys and not in girls, so there is a need for a sex- balanced study to examine potential sex-related differences and moderating effects. Last, most studies have relied on cross-sectional data from offender or high-risk samples; as such, the extent to which previously identified associations generalize to dimensional symptoms in the general population.

Further, the use of data from large, prospective longitudinal studies is needed to clarify the temporal sequence of the association between child neuropsychological functioning and externalizing in adolescence. This is of great importance because it could expand our knowledge on why certain children develop externalising behaviours and others do not. If this is known, prevention strategies could be adapted and optimised.

This study aimed to clarify the role of child neuropsychological functioning as a predictor of later externalising behavioural problems in a large, population-based sample, as well as testing the potential moderating role of callous traits and sex. Using longitudinal data from a large, population-based study, we investigated (a) whether global neuropsychological functioning and its individual domains measured in middle childhood (age 8yrs) prospectively associate with externalising behaviour in emerging adolescence (age 14yrs), and (b) whether callous traits (10 yrs) and sex moderate these relationships. Analyses were adjusted for a range of covariates, as well as baseline externalising behaviour and co-occurring psychopathology to establish the unique links between neuropsychological functioning, callous traits, and later externalising behaviour (the Generation R Study; Kooijman et al., 2016).

We hypothesized that lower neuropsychological functioning in mid childhood would uniquely predict higher externalising behaviour at age 14 over and above existing (i.e. baseline) levels of externalising problems. Second, when looking at the different tasks used to measure neuropsychological functioning, it was expected that tasks indexing EF would not significantly predict externalising behaviour over and above other measured domains, because some studies show multiple domains to be associated with externalising behaviour (Thompson et al., 2019). Third, it was hypothesised that callous traits at the age of 10 years would predict externalising problems at age 14 years (Rizeq et al., 2020). Fourth, based on growing evidence about distinct neurocognitive profiles associated with high- vs low- levels of CU-traits (Dotterer et al., 2021), it is expected that callous traits will significantly moderate the relationship between neuropsychological functioning and externalising problems. Fifth, given that previous studies have focused mainly on males, we expected boys to show more callous traits and externalising, but no a priori hypotheses were formulated regarding sex as a moderator.

The present study is based on data from the Generation R Study, a population-based prospective cohort study (Kooijman et al., 2016), which included pregnant women in Rotterdam, the Netherlands, from 2001–2005. In total, 9778 mothers with a delivery date from April 2002 until January 2006 were enrolled in the study. Data collection of the children started during foetal life and is ongoing. The study is designed to identify early causes leading to normal and abnormal development and health in children throughout their lives. Participants are currently undergoing the 17–19-year assessment wave. The Generation R Study is approved by the Medical Ethical Committee of the Erasmus Medical Center, Rotterdam, the Netherlands.

Children who had data on the measures for neuropsychological functioning, callousness and externalising behaviour were eligible for further analysis. The data used in this study was collected at an average age of 8, 10 and 14 years. Of the 1307 participants who were included in the sub-study on brain imaging and neuropsychological functioning (i.e. where neuropsychological functioning was assessed), 929 participants had data available on externalising problems at mean age 14 years. Descriptive characteristics and percentage of missing data are shown in Table 1. Non-response analyses showed that mothers of children included in this study were higher educated than mothers of non-included children, and they also showed less depression, anxiety, and hostility. Additionally, included children had often a Dutch nationality or a Non-Dutch Western nationality and were younger than excluded children during most data collection moments, except during the assessment of neuropsychological functioning (p < 0.05).

First, descriptive characteristics and correlations between all study variables were calculated. Next, five regression analyses were run. In regression 1, a multiple linear regression analysis was conducted using the total score on the NEPSY-II-NL (i.e. neuropsychological functioning) at age 8 years to predict externalising behaviour at age 14 years. This model was adjusted for baseline externalising problems (age 6), callous traits and all covariates. To test the specificity of associations to externalising problems, the model was also adjusted for internalising problems at age 14 years. In a second regression, multiple linear regression analyses were conducted using different domain scores of the NEPSY-II- NL as simultaneous predictors of externalising behaviour at age 14, adjusting for the same covariates as regression 1. This was done to test whether EF uniquely and significantly predicted externalising behaviour over and above other neuropsychological functioning domains. In regression 3, we examined whether callous traits moderated the association between neuropsychological functioning and subsequent externalising problems. In regression 4, a linear regression analysis was conducted to use neuropsychological functioning at age 8 to predict externalising behaviour at age 14, testing the effects of sex as a moderator. Lastly, a three-way-interaction was examined, i.e. callous traits and sex were both used in a linear regression analysis to see whether the interaction among neuropsychological functioning and callous traits is different across sex.

Graphic visualisations of the interaction effects were performed in R, using the ggplot2 package. In this interaction graph, the association between neuropsychological functioning at age 8 years (x-axis, i.e., the exposure) and externalising problems at age 14 years (y-axis, i.e. the outcome) was plotted separately for children with high versus low callous traits scores. Callous traits scores were dichotomised based on the 93^rd percentile (n = 43 in the high scoring group). Further, externalising problems scores were square root transformed as in the regression models, and both neuropsychological functioning and externalising problems scores were standardised with a mean of zero and a standard deviation of one for comparability.

The results did not support the hypothesis that neuropsychological functioning, or its specific domains such as EF, predict later externalising problems in the general paediatric population, over and above baseline levels of externalising problems. This is not consistent with previous research that have examined the association between subdomains of neuropsychological functioning and externalising behaviour (Blanken et al., 2017; Chow et al., 2018; Petersen et al., 2015). That is, a meta-analysis found a negative relation between language ability and problem behaviour (i.e., Chow et al., 2018). However, this study used only on particular subdomain of neuropsychological functioning, while we aimed at studying the link between more broader externalizing and externalizing behaviour.

Another possible explanation for not finding a relation in our study could be, that in this study we took a more stringent approach by adjusting for multiple factors such as baseline externalising behaviour, co-occurring psychopathology, and individual neuropsychological functioning domains. To our knowledge, there is only one study (using the same sample as the current study), that took baseline externalising behaviour and co-occurring psychopathology into consideration when predicting externalising behaviour through neuropsychological functioning (Blanken et al., 2017). They did find a negative association between externalising and the attention/executive functioning domain of neuropsychological functioning specifically. However, in that study, they examined the relation between neuropsychological functioning and externalising at same age, instead of prospectively as we did in the current study. Our results highlight the importance of taking pre-existing levels of externalising problems into consideration, as this was the strongest predictor for later externalising behaviour in adolescents, consistent with other work showing that externalizing behaviour showed continuity over time (Petersen et al., 2015). It is therefore possible that previous studies would not have found neuropsychological functioning to be significantly predictive of externalising problems when they would have taking baseline externalising problems into account.

Another explanation for the null results regarding neuropsychological functioning as a predictor for externalising behaviour is the large time span between the two measurements. Neuropsychological functioning was measured at 8 years and externalising behaviour at 14 years, corresponding with a mean time span of 5 years between measurements. Studies that did find evidence for neuropsychological functioning as a predictor for later externalising behaviour used a shorter time interval between measurement. For example, in the study of Blanken et al. (2017), the mean time span between two measurements (subdomains of neuropsychological functioning and externalising problems) was 1.85 years. Thompson et al. (2019) measured the two constructs at approximately the same time.

Ultimately, the current results also show that the lack of finding a main affect for neuropsychological functioning might be explained by the moderation of callous traits. That is, the different directions of associations between neuropsychological functioning and externalising depending on the level of callousness, cancel each other out when examining neuropsychological functioning in the whole sample, not stratified by these levels of callousness.

These results are in line with other studies that showed an opposite pattern of associations between neuropsychological functioning and externalising problems based on their levels of callousness (Baskin-Sommers et al., 2015; Dotterer et al., 2021), and also adds to these previous findings in several ways. For instance, Baskin-Sommers et al. (2015) found a divergent relation between executive functions and earlier externalizing problems in predicting violence and substance use in a high-risk sample, while we found this nominally divergent pattern when looking at neuropsychological functioning’to predict externalizing problems in a prospective population-based cohort of child development.

Additionally, although Dotterer et al. (2021) studied intelligence and several aspects of executive functioning, they only found an interaction for CU-traits in sustained attention and antisocial behaviour. However, again, in our sample we additionally found an interaction for overall neuropsychological functioning in a prospective way. The difference might also lie in the fact that this study, as in the Baskin-Sommers et al. (2015) study, the sample was oversampled for children in poverty, and thus high-risk.

Our current findings add to the existing literature by presenting the extended small moderating effect of callous traits in the association between neuropsychological functioning and externalising problems, and in a prospective design.

Based on previous literature, we expected boys to show more callous traits and externalising, but we did not formulate a priori hypotheses regarding sex as a moderator, given that previous studies have focused mainly on males. Current results did show higher levels of externalising and callous traits in males compared to females, but no moderation of sex on the association between neuropsychological functioning and externalising in later adolescence. Although the higher levels of externalising problems and interpersonal callousness in boys was expected and in line with previous findings (e.g., Essau et al., 2006; Mayes et al., 2020; Pihet et al., 2015), the fact that we did not find effects of sex on neuropsychological function is not in line with studies finding sex effects in neuroimaging studies on CU-traits (e.g. Rogers et al., 2019; Villemonteix et al., 2022).

One limitation of the current research is that most participants have the Dutch nationality; consequently, results may not generalise to other national backgrounds.

Another limitation is the use of the shortened NEPSY-II-NL to measure neuropsychological functioning. Due to time constraints, only a few tasks were administered, which means not all neuropsychological constructs are thoroughly measured. Nevertheless, this short version has shown to correlate highly with the total and taps domains of neuropsychological functioning that are previously implicated in externalising psychopathology (see White et al., 2018).

Additionally, the CBCL and interpersonal callousness were measured through maternal reports and not by child reports, which might have led to biased data (Najman et al., 2001). For example, some authors suggest that mothers with elevated psychopathology are more easily distressed by their child’s problem behaviour, resulting in distorted and higher scores on their child’s problems (De Los Reyes & Kazdin, 2005). Therefore, we have adjusted for maternal psychopathology. Furthermore, neuropsychological functioning was objectively assessed by trained researchers.

A last limitation might be that callousness was assessed at age 10 years, so in terms of timing between both predictor (neuropsychological functioning at age 8 years), and outcome measure (i.e. externalising at age 14 years), therefore it was not possible to account for baseline callousness. Consequently, it is unknown what the unique influence is of callousness, over and above pre-existing callousness. Similarly, although externalizing was controlled for at age 6 years, this was not done for age 8 years because there was no data available to include (i.e. was not assessed at that time point). For these reasons, it is recommended to replicate this research, while accounting for baseline callousness.

Last, missing value analyses revealed important differences in neuropsychological function between the excluded participants and included participants. Results showed that excluded participants scored significantly lower on different domains of neuropsychological functioning and covariates. This may have influenced our results. However, multiple imputation was applied to the confounders in order to adjust for potential missingness bias in our analyses.

Children with externalizing problem behaviour are at risk for negative individual functioning later in life. The present findings highlight the need to differentiate between children with externalising problems who also have high levels of callousness and those with low callousness, as they may present with different neuropsychological profiles. Additionally, this differentiation in children and adolescents with problem behaviour might also result in different types of externalizing behaviour. Although not studied here, it might be interesting to study whether children with lower neuropsychological functioning and lower callous traits show other types of externalizing behaviours, such as reactive forms of aggression, while those children with higher neuropsychological functioning and high callousness are inclined to show more proactive forms of violence and cruel behaviour (Rosa-Justicia et al., 2020). Although the current study added on evidence suggesting the delineation of two subgroups by examining associations between neuropsychological functioning, externalising behaviour and callousness prospectively, the mechanisms of this interactive relation remain unclear. Causal relations might need to be further established using causal inference methods.

Current findings also support new directions in clinical practice, where a distinction in different groups of externalizing problem behaviour characterized by high versus low levels of CU- traits has recently been adopted by the DSM-5 through the inclusion of a ‘limited-prosocial emotions’ specifier for CD, with these two subgroups showing distinct profiles in terms of aetiology, behavioural outcomes and neurocognition (Vanwoerden et al., 2016; Viding & McCrory, 2018).

This specification of children with conduct problems and differing levels of CU-traits has also implications for treatment of these problems (Andrade et al., 2022). That is, since they show different core dysfunctions underlying their behaviour, treatment must match with these different core dysfunctions by either focusing more on the low neuropsychological functioning or focus more on the callous traits. However, more research is needed to determine which treatment intervention works best for children displaying externalising behaviour with varying levels of neuropsychological functioning, and for those with and without CU-traits (Bakker et al., 2017).

In sum, the current findings do not show support for the association between child neuropsychological functioning and externalising behaviour in preadolescence within the general population. However, when stratifying by callous traits, results indicate that neuropsychological functioning does associate with externalising problems, but in differential directions depending on levels of callousness. That is, in children who display higher levels of callousness, higher neuropsychological functioning relates to externalising behaviour in later adolescence, whereas in those with lower levels of callousness, no association was found. This study also highlights the importance of child-onset problem behaviour and how this predicts later externalising problems in adolescence. Treatment efforts should be directed at preventing externalising problems at a young age, with the emphasis on children who display high CU-traits. Research has shown that if interventions are tailored fitting the unique emotional, cognitive, and motivational profile of children with CU-traits, behavioural problems can be reduced (Frick et al., 2014). Future studies should look further into the underlying neurocognitive mechanisms of why elevated CU-traits and high neuropsychological functioning lead to more externalising behaviour and aggression in later life.