Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 1.
Published in final edited form as: J Abnorm Psychol. 2020 May 28;129(6):633–645. doi: 10.1037/abn0000525

Evidence for two genetically-distinct pathways to co-occurring internalizing and externalizing problems in adolescence characterized by negative affectivity or behavioral inhibition

Frances L Wang 1, Chardée A Galán 1, Kathryn Lemery-Chalfant 2, Melvin N Wilson 3, Daniel S Shaw 1
PMCID: PMC7415528  NIHMSID: NIHMS1590071  PMID: 32463263

Abstract

Unique pathways to adolescents’ co-occurring internalizing/externalizing problems, a severe and common form of psychopathology, remain poorly delineated; this paucity of knowledge impedes the development of personalized interventions. We examined established measures of genetic risk and early childhood temperamental dimensions to clarify potentially distinct pathways to adolescents’ co-occurring internalizing/externalizing problems. Participants were drawn from a longitudinal randomized controlled trial of a family-based intervention. The study employed multiple informants and methods, including observer ratings of toddlers’ negative affectivity and behavioral inhibition, and primary caregiver ratings of toddlers’ inhibitory control; internalizing and aggression polygenic risk scores (PRS) based on prior meta-GWAS’; and parents’ and teachers’ reports of adolescents’ internalizing and externalizing problems. Higher levels of the aggression PRS indirectly predicted primary caregiver- and teacher-reported co-occurring problems relative to all other groups through greater early childhood negative affectivity. Lower levels of the aggression PRS and higher levels of the internalizing PRS indirectly predicted co-occurring problems relative to the externalizing ‘only’ and low problem groups (primary caregivers only) through greater early childhood behavioral inhibition. Findings suggest two different genetic pathways to co-occurring problems that could lead to distinct prevention and intervention efforts.

Keywords: Genetics, Polygenic Risk Scores, Co-occurring Internalizing and Externalizing, Adolescence, Temperament

Adolescents who exhibit both internalizing and externalizing problems are diverse with respect to their underlying genetics. Despite exhibiting the same symptom presentation, adolescents’ unique etiologies suggest that distinct interventions may be needed.

The co-occurrence between internalizing and externalizing problems is frequent in adolescence (Caron & Rutter, 1991) and portends worsened psychosocial outcomes concurrently and into adulthood (Capaldi, 1992; Diamantopoulou et al., 2011). Extant theories support the notion that there exist several unique pathways to co-occurring problems in adolescence (Caron & Rutter, 1991), consistent with models of equifinality (Cicchetti & Rogosch, 1996). Equifinality also suggests that different prevention and intervention strategies may be necessary for youths whose co-occurring problems have distinct etiologies. Although distinct pathways to adolescents’ co-occurring problems are poorly understood, youths’ genetic susceptibilities and early temperamental propensities may help delineate them. Moreover, these pathways are important to study with regards to adolescents’ outcomes, as this period marks a sensitive stage of development that may be particularly influential on adulthood adjustment (Dahl et al., 2018).

Causes of Internalizing and Externalizing Co-occurrence

There exist several hypotheses about the causes of co-occurrence between internalizing and externalizing problems, most with some degree of empirical support (see Oland & Shaw, 2005 for a review). The causal hypothesis is one of the most widely studied, and indeed, research has shown that externalizing problems prospectively predicted risk for internalizing problems (e.g., Moilanen, Shaw, & Maxwell, 2010) and vice versa (e.g., Loeber & Keenan, 1994). In addition, a substantial body of work has shown that internalizing/externalizing co-occurrence may be due to shared vulnerabilities or risks for both problems, such as genetic and environmental influences (Oland & Shaw, 2005). For example, a common genetic component explained between 45% (O’Connor et al., 1998) to 62% (Cosgrove et al., 2011) of the covariation between internalizing and externalizing problems. A common genetic factor was also found to influence disorders across internalizing and externalizing domains (Lahey et al., 2011; Tackett et al., 2013). Moreover, the heritability of a general factor of psychopathology, which reflects the common variance between internalizing and externalizing problems and is derived from a bifactor model, ranged from 38–63% across molecular and behavioral genetic methods (Grotzinger et al., 2019; Neumann et al., 2016). Based on the substantial role of genetics, this study focused on genetic influences on co-occurring internalizing/externalizing problems.

We operationalized co-occurring problems using latent profile analysis (LPAs), which allows comparison of a latent co-occurring problems ‘group’ relative to an internalizing ‘only,’ externalizing ‘only,’ and low problem behavior group. In addition, in a secondary, complementary approach, we used a bifactor model to operationalize co-occurring problems. The bifactor approach yields three dimensional latent factors, including a ‘general’ or co-occurring problems factor (on which all internalizing and externalizing indicators load), a ‘specific’ externalizing problems factor (externalizing indicators only), and a ‘specific’ internalizing problems factor (internalizing indicators only). As all three factors are typically specified to be orthogonal (Chen et al., 2006), co-occurrence could be observed in multiple ways, such as if one individual scored highly on the co-occurring factor only, whereas another scored highly on both internalizing and externalizing factors only. Thus, although LPA analyses may be more straightforward in elucidating a broad group showing co-occurring problems, bifactor analyses may provide more insights into mechanisms underlying co-occurring problems.

Roles of Temperamental Traits

Dimensions of temperament are defined as constitutionally-based individual differences in reactivity and self-regulation (Rothbart et al., 2001). Rothbart’s model of children’s temperament suggests organization into three broad domains, including effortful control (i.e., ability to suppress a dominant response to perform a subdominant response), surgency (i.e., approach- and reward-orientation, positive affect), and negative affectivity (i.e., propensity towards negative emotions such as frustration, fear, and sadness; Rothbart, Ahadi, Hershey, & Fisher, 2001).

Temperamental traits have been conceptualized as core underlying features explaining psychological comorbidity. For example, in cross-sectional analyses, trait negative emotionality, surgency-related traits, like impulsivity and behavioral inhibition, and effortful control have been shown to play a role in the manifestation of both internalizing and externalizing problems in childhood (Hankin et al., 2017; Olino et al., 2014; Shields et al., 2019; Tackett et al., 2013). Although complex transactional relations with environmental and other child factors influence longitudinal associations between dimensions of temperament and problem behavior (Sanson et al., 2004), previous studies have also found direct, longitudinal effects of temperament on co-occurring problems in childhood and adolescence, such as early childhood behavioral inhibition (Edwards & Hans, 2015), difficultness/irritability (Fanti & Henrich, 2010), and middle childhood effortful control (Wang et al., 2016). Thus, all three temperamental dimensions may partially explain the co-occurrence of internalizing and externalizing problem behaviors.

Moreover, because temperamental traits have been shown to be heritable (Goldsmith et al., 1997), they may partially explain why internalizing and externalizing problems share genetic influences. Prior studies showed that a common genetic factor accounted for the covariation among negative affectivity, internalizing, and externalizing problems in 9–13 year-olds (Mikolajewski et al., 2013), 12–18 year-olds (Hink et al., 2013) and 9–17 year-old boys (Tackett et al., 2011). Genetic influences on novelty seeking, a lower-order facet of surgency, also partially accounted for the genetic covariation between depression and externalizing problems in 12–18 year olds (Hink et al., 2013). Tackett et al. (2013) found that a general psychopathology factor was the most strongly genetically-correlated with negative emotionality, followed by daring and prosociality, in 9–17 year olds. Finally, Lemery-Chalfant, Doelger, and Goldsmith (2008) found that effortful control overlapped genetically with either internalizing or externalizing problems separately in middle childhood. In this study, we examined how three early childhood temperament dimensions were genetically linked with co-occurring internalizing/externalizing problems. We examined inhibitory control (i.e., facet of effortful control involving the control of overt behaviors), negative affectivity, and behavioral inhibition (i.e., low surgency, high fearfulness).

Are There Multiple Genetic and Temperamental Pathways to Co-occurring Problems?

Prior studies significantly enhanced our understanding of the etiology of psychopathology, and particularly co-occurring internalizing/externalizing problems. Unfortunately, they have not disentangled whether varied genetic origins exist that lead to co-occurring problems in adolescence. Thus, it remains unclear whether the shared genetic variance underlying internalizing and externalizing problems and one temperamental dimension is orthogonal to that shared by internalizing and externalizing problems and another temperamental dimension. Such a finding would suggest that there are two (or more) genetic diatheses to co-occurring problems stemming from different temperament dimensions. This issue is important to address, as different genetic pathways to co-occurring problems would suggest distinct etiologies and, potentially, different prevention and intervention strategies.

One way to address whether there are varied genetic origins leading to co-occurring problems is through the use of polygenic risk scores (PRS). PRSs represent the aggregate effect of many common single nucleotide polymorphisms (SNPs) weighted based on each SNP’s effect on an outcome in a genome-wide association study (GWAS). Pertinent to this study are two large GWAS on childhood internalizing problems or aggression1 (Benke et al., 2014; Pappa et al., 2016), which can be used to create PRSs (Elam et al., 2018). It is advantageous that these PRSs are expected to represent broad genetic risk for internalizing problems or aggression, as opposed to more specific biological mechanisms, because it suggests that they should encompass genetic risk shared with different dimensions of temperament.

Examining interrelations among PRSs, temperament, and adolescents’ co-occurring problems could help determine whether there exist distinct genetic pathways to co-occurring problems. For example, temperamental dimensions that each predict LPA-defined co-occurring problems may be predicted by different PRSs, suggesting two distinct genetic pathways to co-occurring internalizing/externalizing problems. Moreover, if one of these genetically-distinct temperamental dimensions predicted the co-occurring factor of the bifactor model, whereas another predicted the ‘specific’ internalizing and externalizing factors, this pattern of findings would further cement mechanistically different kinds of co-occurring problems. Alternatively, prior research demonstrated that many temperament dimensions showed small-to-moderate genetic intercorrelations (Gillespie et al., 2003). Thus, several temperament dimensions that each predict co-occurring problems could each be predicted by the same PRS, suggesting that co-occurring problems stem from a broad genetic diathesis shared by multiple temperamental difficulties.

How Early Can Genetic Risk Pathways be Identified?

Most studies on shared genetic influences underlying childhood temperament and behavior problems were conducted cross-sectionally. It remains less clear how early we can identify and, thus, intervene with youth following particularly risky genetic pathways to co-occurring problems in adolescence. Indeed, note that genetically-influenced traits have been shown to be amenable to environmental interventions (Lemery-Chalfant et al., 2018). Unfortunately, no genetically-informative studies to our knowledge have tested prospective effects of temperament dimensions as early as ages 2 and 3 on adolescents’ co-occurring problems over a 12-year span. In this study, we examined how PRSs for internalizing problems or aggression predicted early childhood inhibitory control, negative affectivity, and/or behavioral inhibition at child ages 2 and 3, and in turn, how these dimensions of temperament predicted co-occurring internalizing/externalizing problems at age 14, relative to adolescents without such problems. Because we were interested in early genetic risk, we used data from the early childhood samples of internalizing and aggression GWASs (mean ages 2–5; Benke et al., 2014; Pappa et al., 2016).

We tested mediating influences of the three temperament dimensions at both ages 2 and 3, within the same model, as each age could be important for differing reasons. Indeed, the current sample is drawn from a larger randomized controlled trial beginning after the age 2 assessment, so age 2 temperament would importantly not be influenced by the child’s intervention group status. However, as temperament also develops rapidly across early childhood (Lemery et al., 1999) and by age 3 has been shown to be predictive of adult psychopathology (Caspi et al., 1996), age 3 temperament dimensions could be better predictors of adolescents’ co-occurring problems. Although temperament dimensions in early childhood are only modestly-to-moderately stable, particularly effortful control (Kochanska et al., 2000) and observational measures of temperament (Majdandžić & Boom, 2007), we nonetheless also assessed whether it was feasible to aggregate temperament scores at ages 2 and 3.

Current Study

We tested whether PRSs for early childhood internalizing problems or aggression predicted a higher likelihood of membership in a co-occurring internalizing/externalizing problems group at age 14 relative to internalizing ‘only,’ externalizing ‘only,’ and low externalizing and internalizing problem groups. Problem behavior groups were identified using LPAs of parent- and teacher-reported internalizing and externalizing problems, previously reported in another study of this sample focused on parental mediators of intervention effects (Wang et al., 2019). As prior research showed that a trajectory group characterized by longitudinally co-occurring conduct and emotional problems in adolescence showed the highest genetic risk relative to other groups (Hannigan et al., 2018), we hypothesized that the co-occurring group would show higher aggression and internalizing PRSs than internalizing ‘only’, externalizing ‘only’ and low problems groups. We also tested a secondary, complementary approach of using as an outcome a primary caregiver-reported bifactor model. No hypotheses about genetic risk associations were formed based on the lack of literature comparing the heritability of the general to ‘specific’ factors.

We tested whether early childhood behavioral inhibition, negative affectivity, or inhibitory control mediated the relation between the internalizing or aggression PRSs and adolescents’ LPA-derived co-occurring problem behaviors or bifactor model factors; age 2 and 3 temperament dimensions and their stabilities were tested (or aggregate measures over time if stabilities were high). Although it was unclear how PRSs would predict different dimensions of temperament, we tentatively hypothesized that greater internalizing PRSs would predict greater behavioral inhibition, greater aggression PRSs would predict lower inhibitory control, and greater internalizing and aggression PRSs would predict greater negative affectivity. Based on prior studies, we hypothesized that adolescents in the co-occurring LPA group would exhibit lower early childhood behavioral inhibition and inhibitory control, as well as higher negative affectivity. We also hypothesized that the co-occurring latent factor (bifactor model) would be predicted by lower inhibitory control and higher negative affectivity and behavioral inhibition, the specific internalizing factor would be predicted by higher behavioral inhibition and negative affectivity, and the specific externalizing factor would be predicted by lower inhibitory control and higher negative affectivity. In sum, we hypothesized two genetically-distinct pathways to co-occurring problems that operate through either behavioral inhibition or inhibitory control, and that each of these distinct pathways would show features of negative affectivity. Moreover, we hypothesized that each PRS would result in co-occurrence by exerting an influence on temperament dimensions that are shared between internalizing and externalizing problems (i.e., the ‘general factor’), suggesting that among the ‘shared’ variance between internalizing and externalizing problems, some portions are nonetheless distinct from one another genetically.

Study goals were tested using data from a large longitudinal sample of low-income and racially diverse youth. Half were randomized to a nearly-annually-delivered family-based preventive intervention (the Family Check-Up [FCU]) or a control group receiving nutritional services as usual as part of the Women, Infants, and Children Nutritional Supplement Program (WIC).

Methods

Participants

Participants included 731 caregiver-child dyads recruited between 2002–2003 from WIC programs providing nutritional assistance for impoverished families in and around Pittsburgh, PA, Eugene, OR, and Charlottesville, VA (Dishion et al., 2008). Families were approached at WIC sites and invited to participate if they had a child between age 2 years 0 months and 2 years 11 months of age. Families were eligible if they scored at least one SD above the normative mean in two domains: a) familial (i.e., maternal depression, teen parent when first child was born, daily parenting challenges, parent substance use or mental health diagnosis), b) sociodemographic (i.e., low parental education, low family income), or c) child (i.e., conduct problems, high-conflict relationship).

Recruitment

Of the 1,666 families with children in the age range, 879 were eligible. Of the 731 families (49% female; Mage=29.9 months, SDage=3.2) that consented to participate, 272 were recruited in Pittsburgh, 271 in Eugene, and 188 in Charlottesville. Before the first home assessment at age 2, 367 of the 731 enrolled families were randomly assigned to the FCU intervention and 364 families were assigned to the control condition (WIC services as usual). Primary caregivers self-identified as European American (50%), African American (28%), biracial (13%), and other groups (9%; e.g., American Indian, Native Hawaiian). Primary caregivers participating in age 2 assessment tasks were predominantly biological mothers (96%). More than two-thirds of enrolled families reported an annual income less than $20,000.

Procedures

Primary caregivers and children participated in yearly assessments conducted at home from ages 2 to 5, 7.5 to 10.5, and then at age 14. Parents completed questionnaires regarding socio-demographics, parenting, neighborhood conditions, and child behavior. Parent-child observation tasks were conducted during age 2 and 3 assessments, with all tasks videotaped. Parental written consent was obtained and assent from children beginning at age 14. Participants were compensated $100 during the age 2 assessment, $120 at the age 3 assessment, and $210 at the age 14 assessment. Institutional Review Boards at each site approved study protocols.

Participants provided saliva samples with Oragene kits for genotyping. RUCDR Infinite Biologics at Rutgers University extracted and normalized the DNA. Samples were genotyped using the Affymetrix Axiom Biobank Array. 4,098,692 SNPs remained after basic post-imputation data cleaning. SNPs not in Hardy-Weinberg equilibrium (p<10−6), with a minor allele frequency less than 1%, or those that fell within the Major Histocompatibility Complex on Chromosome 6 were removed. Copy-number variations were removed if they did not meet the 5% missing gene data threshold. 4,048,277 SNPs remained in imputed data after quality control procedures. Using the sliding window procedure in PLINK, we reduced linkage disequilibrium (LD) by screening out regions of long-range and local LD.

Current sample

Adolescents genotyped at age 14 comprise the current subsample (n=515 or 86.7% who participated in age 14 home visits; 129 from VA, 184 from OR, and 202 from PA). Adolescents were 50% female and belonged to the following racial/ethnic groups: 10% Latino, 30% African American, 48% European American, 5% Native American, 1% Asian American, and 6% other or unknown race.

Selective attrition analyses revealed no significant differences between those who did not vs. did provide a saliva sample during the age 14 assessment with respect to parental education (high school diploma vs. not), χ2(1) =1.40, p=0.24; minority racial status (White vs. non-White), χ2(1) = 0.04, p=0.84; sex of child, χ2(1) = 0.67, p=0.41; intervention status, χ2(1)=0.11, p=0.74; study site, χ2(2)=0.49, p=0.78; maternal depression at age 2, t(543)= −0.29, p=0.77; behavioral inhibition at age 2, t(524)= −0.09, p=0.93, or at age 3, t(478)= −0.19, p=0.84; inhibitory control at age 2, t(546)= −1.19, p=0.24, or at age 3, t(521)= 0.66, p=0.51; negative affectivity at age 2, t(552)= −1.46, p=0.15, or at age 3, t(503)= 0.07, p=0.95; internalizing at age 2, t(553)=−0.51, p<0.61; or externalizing at age 2, t(553)=−1.05, p<0.29.

Measures

Behavioral inhibition

At ages 2 and 3, child behavioral inhibition was coded based on reactions to an approach by an adult stranger (2 minutes) and two novel objects (2 minutes each): a tunnel and a mechanically-operated robot. Based on a system developed by (Kochanska, 1991a), coders assigned one global rating for the child’s inhibition on all three tasks on a 4-point scale ranging from 1 (not inhibited) to 4 (much inhibited). Coders scored 20% of the tapes as a team and inter-rater reliability Kappa coefficients ranged between .68-.83.

Inhibitory control

Primary caregivers reported on 2 and 3 year olds’ inhibitory control over the past six months using 13-items (7-point scale) on the inhibitory control scale of the Children’s Behavior Questionnaire (Rothbart et al., 2001; αage2=0.68, αage3=0.69).

Negative and positive affect

At child ages 2 and 3, children and their primary caregivers participated in parent-child joint play (12 mins), clean up (5 min), and meal (20 mins) observational tasks. Interactions were coded using the Relationship Affect Coding System (RACS; Peterson et al., 2008), a micro-social coding system assessing verbal, physical, and affect characteristics for each family participant. The current study focused on affect codes. Coding was carried out using the software, Noldus Observer XT, v.11.0 (Noldus, 2012), which allows continuous coding of an interaction between child and caregiver. Anger, disgust, distress, and ignoring were categorized as negative emotions, and validation and positive affect as positive emotions. Across tasks, coders showed 72% agreement (Kappa=0.68) at age 2 and 94% agreement (Kappa=0.93) at age 3. Children’s negative affect score was a mean composite of negative affect during tasks. Primary caregivers’ negative and positive affect scores were averages of affect during tasks (including behavioral inhibition) and were included as covariates to rule out parents’ task-related affect as confounding variables.

Internalizing and externalizing problems

Primary caregivers rated 14-year-olds’ problem behaviors using the Child Behavior Checklist for ages 6–18 (CBCL; Achenbach & Rescorla, 2001) using a 3-point scale, from 0 (not true) to 2 (very true or often true). Subscales used for analyses included withdrawn/depressed (α=.83), anxious/depressed (α=.84), somatic problems (α=.79), aggressive behaviors (α=.93), rule-breaking (α=.82), and attention problems (α=.88). Teachers rated the subscales from the Teacher Report Form (Achenbach & Rescorla, 2001) at age 14: withdrawn/depressed (α=.86), anxious/depressed (α=.84), somatic problems (α=.74), aggressive behaviors (α=.96), rule-breaking (α=.77), and attention problems (α=.94). Questionnaires were not sent to teachers until they had known the child for at least two months.

Aggression and internalizing PRSs

PRSs were created using PRSice v2 (Euesden et al., 2015) and PLINK v1.9 (Purcell et al., 2007). The aggression PRS was based on the early childhood sample in an aggression meta-GWAS (Pappa et al., 2016) and the internalizing PRS on a CBCL internalizing problems meta-GWAS in preschoolers (Benke et al., 2014). Of our 4,048,277 SNPs, 1,544,569 overlapped with the aggression meta-GWAS and 1,478,114 with the internalizing meta-GWAS. Synonymous SNPs were removed. PLINK’s clumping procedure accounted for non-independence among SNPs (threshold of r2=.3 and 250 kb), resulting in 172,038 independent SNPs to potentially include in the aggression PRS and 168,631 SNPs for the internalizing PRS. See Supplement for more detail on data cleaning.

Four aggression PRSs were created by unit-weighting SNPs below p-value thresholds of .001, .01, .05, and .10 (effect weights not available for the early childhood sample). Four internalizing PRSs were created by effect-weighting SNPs below those four p-value thresholds. Researchers have recommended the optimal threshold for PRSs to be the strictest p-value that maximizes explained variance (Evans et al., 2013). The only significant correlation between PRSs and age 2 and 3 problem behaviors was between the p < 0.001 aggression PRS and age 3 externalizing problems. Because there was no clear ‘optimal’ PRSs, we chose p<0.001 because it was the most stringent and may reduce noise and false positives. Final aggression and internalizing PRSs included 388 and 447 SNPs, respectively.

Other covariates

Primary caregivers reported on the sex of the child (0 = male, 1 = female), their levels of education (1 (no formal schooling) to 9 (graduate degree)), and 2-year-olds’ internalizing (α=.81) and externalizing problems (α=.86; CBCL 1 ½ - 5; Achenbach & Rescorla, 2000). At child age 2, mothers reported their depressive symptoms during the past week using the Center for Epidemiological Studies of Depression Scale (α=.88; Radloff, 1977). Items were summed. The urbanicity of each study site (1=rural, 2=suburban, 3=urban) and intervention status were covariates.

To account for population admixture, we conducted principal components analysis of all autosomal SNPs. PLINK was used to extract the first 20 components. The first component (PC1) had an eigenvalue of 28.84 and differentiated European-American and Latino from African-American groups, with most biracial participants falling in the middle. The second component (PC2) had an eigenvalue of 5.62 and differentiated non-Latino (European and African American) from Latino participants. Remaining components had small eigenvalues (1.21–1.45) and were excluded.

Analytic Plan

Analyses were conducted in Mplus version 7.11 using maximum likelihood with robust standard errors and full information maximum likelihood for missing data estimation (Muthén & Muthén, 1998). Groups with distinct patterns of internalizing and externalizing problems were identified using latent profile analyses (LPA). Note, LPAs and resultant groups were previously conducted and detailed in a prior study of this sample on intervention effects (Wang et al., 2019). For clarity, we reiterate the procedures here. LPAs were conducted separately for primary caregivers and teachers using as indicators CBCL or TRF T-score subscales. One-to-five group models were compared on Bayesian Information Criteria (BIC), adjusted BIC, bootstrapped likelihood ratio test (BLRT) and log likelihoods (Nylund et al., 2007). We evaluated whether each new group added distinct, meaningful information about the sample population. Entropy, a measure of classification certainty, was evaluated to determine whether it was appropriate to save the ‘most likely class membership’ as a manifest variable.

We modeled a bifactor model by specifying three, orthogonal latent factors. All externalizing and internalizing subscales loaded onto a general psychopathology factor, externalizing scales onto the externalizing factor, and internalizing scales on the internalizing factor. Model fit was determined using conventional fit statistics (chi-square, CFI, RMSEA).

Multinomial logistic regression was conducted to test predictors of adolescents’ latent profiles. Linear regression tested predictors of the bifactor model. These models were conducted separately, although the following procedures were the same for both. Continuous predictors and covariates were standardized. Adolescents’ group membership (or latent factors of psychopathology) were regressed on predictors (internalizing and aggression PRSs), mediators (age 2 and 3 temperament), and covariates (age 2 internalizing and externalizing, intervention, sex of child, urbanicity, parental education, ancestry, maternal depression, age 2 and 3 parental positive and negative affect). Temperament dimensions were regressed on both PRSs and all covariates, except for intervention status (for age 2 temperament) and age 2 problem behaviors. Autoregressive paths were modeled for temperament over time. Correlations were modeled among: (1) age 2 mediators, internalizing, and externalizing, and (2) age 3 mediators.

Planned contrasts for latent profiles included: (1) co-occurring problems (reference group, coded 0) compared to each group and, (2) low problems (reference group, coded 0) compared to internalizing and externalizing ‘only’ groups. The joint significance test was used to test mediation, in which a mediated effect is significant if both ‘a’ and ‘b’ paths are significant. This approach has been shown to provide the best balance of Type I error and power against a range of other methods (MacKinnon et al., 2002).

Results

Zero-order Correlations

See Table 1. Higher internalizing and aggression PRSs were related to higher and lower levels, respectively, of behavioral inhibition (child age 2). Higher aggression PRS was significantly correlated with lower inhibitory control (age 3). Girls showed higher behavioral inhibition, inhibitory control, and lower negative affectivity than did boys (age 2), as well as lower externalizing problems (age 3). Children from more urban communities showed higher internalizing and aggression PRSs. As the stability of behavioral inhibition and negative affectivity were below r=0.25, we decided against combining age 2 and 3 temperament. Internalizing and aggression PRSs were not correlated. See supplementary Table 3 for zero-order correlations with indicators of latent profiles.

Table 1.

Zero-order Correlations

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.
1. INT PRS 1
2. AGG PRS 0.03 1
3. PC1 0.33** 0.05 1
4. PC2 −0.02 0.18** 0.000 1
5. Sex of child −0.01 −0.02 0.01 −0.04 1
6. Urbanicity 0.09* 0.12** 0.17** 0.19** 0.01 1
7. Parent Education −0.02 0.02 −0.04 0.28** −0.03 0.14** 1
8. BI (2) 0.09* −0.16** −0.03 −0.13** 0.09* −0.01 −0.05 1
9. NA (2) −0.02 0.02 −0.002 0.003 −0.10** −0.01 0.02 0.07 1
10. IC (2) 0.02 −0.08 0.07 −0.06 0.12** 0.02 0.02 −0.01 −0.10** 1
11. BI (3) 0.02 0.02 −0.02 −0.06 0.08 0.04 −0.07 0.24** −0.08* 0.001 1
12. NA (3) −0.05 0.07 0.01 −0.02 −0.04 −0.01 0.04 0.07 0.16** −0.05 −0.01 1
13. IC (3) 0.03 −0.10* −0.03 −0.09* 0.14** −0.05 0.13** 0.03 −0.05 0.51** 0.06 −0.09* 1
14. INT (2) 0.07 −0.01 0.10* −0.16** 0.04 0.01 −0.19** 0.06 0.06 −0.19** 0.05 0.08* −0.18** 1
15. EXT (2) −0.01 0.002 −0.01 0.10* −0.06 0.08* −0.08* −0.03 0.08* −0.49** −0.01 0.04 −0.42** 0.52** 1
Open in a new tab
**

p<0.01.

*

p<0.05.

INT: Internalizing. EXT: Externalizing. PRS: Polygenic Risk Score. PC: Principal Component. BI: Behavioral Inhibition. NA: Negative Affectivity. IC: Inhibitory Control. Sex: 0=male, 1=female. Urbanicity: 0=rural, 1=suburban, 2=urban. (2) or (3) indicates measurement at age 2 or 3.

Latent Profile Analyses

Fit indices improved with the addition of groups. For both reporters, the five-group models showed minimal improvements in fit over the four-group model (see supplementary Table 1). Each five-group model extracted a group containing <5% of the sample and did not contain substantively different groups from the four-group model. Four-group models were selected, including, (1) low problems, (2) elevated internalizing relative to externalizing (hereafter referred to as internalizing problems ‘only’), (3) elevated externalizing relative to internalizing (i.e., externalizing problems ‘only’), and (4) co-occurring problems. Final LPAs showed high entropy, indicating high classification certainty (0.90–0.93). Adolescents’ ‘most likely class membership’ was analyzed as a manifest variable. See supplementary Table 2 and Figure 1 for visual depiction of final LPAs, mean subscale T-scores for each group, and group percentages. Those in co-occurring groups scored above the clinical (70) or borderline clinical (65) cut-off on all subscales except teacher-reported somatic complaints and attentional problems. Those in internalizing and externalizing ‘only’ groups scored at or just below (>64) the borderline clinical cut-off for 75% of their respective scales (supplementary Table 2).

Figure 1.

Figure 1.

Open in a new tab

Final model results. Covariates, correlations among mediators within-age, non-significant paths, and direct effects from PRSs to age 14 outcomes are not shown to enhance interpretability. CO: co-occurring. INT: Internalizing ‘only’. EXT: Externalizing ‘only’. Low: Low problems. “Co > Low” denotes that co-occurring group was higher on that variable than low problems group. Standardized coefficients depicted (*p<0.05). Akaike Information Criteria=28190.511, Bayesian Information Criteria=29451.029, Sample-Size Adjusted Bayesian Information Criteria=28508.298.

Bifactor Model

The bifactor model fit the data well: χ2(3)=3.52, p=0.32, RMSEA=0.02, CFI=0.99. Standardized factor loadings ranged from 0.48–0.78 (p<0.001) for the general factor, 0.17–0.47 (p<0.045) for the externalizing factor, and 0.30–0.55 (p<0.001) for the internalizing factor.

PRS Prediction of Temperament

See Tables 2 and 3 for full results and Figure 1 for path models. Marginally significant effects on latent profiles were only described in the text below if the corresponding effect was significant with the second reporter.

Table 2.

Prediction of Mediators

Mediators age 2 Mediators age 3
Behavioral Inhibition Negative Affectivity Inhibitory Control Behavioral Inhibition Negative Affectivity Inhibitory Control
Internalizing Polygenic Risk Score 0.11(0.04)* −0.03(0.04) −0.01(0.05) −0.003(0.05) −0.04(0.04) 0.05(0.04)
Aggression Polygenic Risk Score −0.13(0.04)* 0.02(0.04) −0.08(0.04)† 0.08(0.04)† 0.09(0.04)* −0.02(0.04)
Urbanicity −0.02(0.05) −0.03(0.05) 0.02(0.04) 0.03(0.05) −0.04(0.04) −0.05(0.04)
Parental Education −0.03(0.05) 0.01(0.05) 0.04(0.04) −0.03(0.05) 0.06(0.04) 0.13(0.04)*
Ancestry Principal Component 1 −0.06(0.05) −0.003(0.04) 0.08(0.05)† −0.03(0.05) 0.03(0.04) −0.06(0.04)
Ancestry Principal Component 2 −0.10(0.06)† −0.01(0.05) −0.05(0.04) −0.04(0.04) −0.04(0.04) −0.07(0.04)*
Sex of Child 0.04(0.04) −0.03(0.04)† 0.10(0.04)* 0.03(0.05) 0.04(0.04) 0.02(0.04)*
Parent Negative Affect Age 2 0.02(0.05) 0.28(0.06)** −0.05(0.03) 0.04(0.04) 0.36(0.14)* 0.01(0.02)
Parent Positive Affect Age 2 0.11(0.06)† 0.15(0.10) −0.02(0.04) 0.07(0.04) 0.11(0.05)* 0.05(0.04)
Maternal Depression Age 2 −0.02(0.04) −0.04(0.04) −0.02(0.05) 0.02(0.04) −0.04(0.04) −0.10(0.05)*
Intervention -- -- -- −0.01(0.05) −0.02(0.04) 0.06(0.04)
Behavioral Inhibition Age 2 -- -- -- 0.27(0.06)** 0.05(0.04) 0.01(0.04)
Negative Affectivity Age 2 -- -- -- −0.09(0.06)* 0.19(0.08)* −0.002(0.07)
Inhibitory Control Age 2 -- -- -- 0.04(0.05) 0.04(0.05) 0.49(0.04)**
Open in a new tab

Notes. n=515.

**

p<0.001

*

p<0.05

†

p<0.10.

Standardized Coefficients(Standard Errors) presented. Effects of predictors on mediators were identical across parent and teacher models, they are only presented once. Sex of child: 0=male, 1=female. Urbanicity: 0=rural, 1=suburban, 2=urban. Intervention: 0=control, 1=intervention.

Table 3.

Prediction of Adolescents’ Problem Behaviors

Primary Caregiver-Reported Outcomes Teacher-Reported Outcomes
CO vs. Low CO vs. INT CO vs. EXT Low vs.INT Low vs.EXT CO vs. Low CO vs. INT CO vs. EXT Low vs. INT Low vs. EXT
Internalizing PRS −0.09(0.09) −0.24(0.12)* −0.26(0.12)*   −0.38(0.14)* −0.39(0.13)* 0.23(0.15) 0.07(0.15) 0.17(0.15) −0.32(0.18)† −0.09(0.14)
Aggression PRS −0.02(0.09) 0.03(0.10) 0.01(0.10) 0.15(0.14) 0.08(0.13) −0.10(0.14) −0.07(0.15) −0.03(0.15) 0.05(0.17) 0.13(0.13)
BI age 2 −0.16(0.08)* −0.17(0.10)† −0.20(0.09)* 0.03(0.15) −0.04(0.16) −0.16(0.11) 0.01(0.12) −0.20(0.12)† 0.37(0.19)* −0.10(0.17)
NA age 2 0.13(0.08) 0.14(0.11)† 0.14(0.10) −0.03 (0.14) −0.03(0.15) −0.05(0.12) −0.15(0.13) −0.08(0.13) −0.26(0.20) −0.08(0.16)
IC age 2 −0.007(0.08) −0.03(0.10) 0.02(0.10) −0.10(0.18) 0.02(0.17) 0.15(0.15) 0.15(0.16) 0.18(0.16) 0.04(0.23) 0.06(0.18)
BI age 3 −0.04(0.08) −0.04(0.10) −0.01(0.10) 0.01(0.15) 0.07(0.15) −0.004(0.13) −0.03(0.14) 0.01(0.14) −0.05(0.21) 0.03(0.17)
NA age 3 −0.18(0.06)* −0.21(0.07)* −0.25(0.09)* −0.004(0.16) −0.09(0.17) −0.33(0.08)** −0.25(0.10)* −0.37(0.10)** 0.13(0.18) −0.11(0.17)
IC age 3 0.05(0.08) 0.11(0.10) 0.01(0.10) 0.14(0.16) −0.11(0.16) 0.03(0.17) −0.04(0.18) −0.07(0.17) −0.14(0.24) −0.18(0.17)
Urbanicity −0.10(0.08) −0.14(0.10) −0.14(0.10) −0.07(0.15) −0.07(0.14) −0.05(0.14) 0.11(0.14) −0.02(0.14) 0.37(0.20)† 0.07(0.14)
Parental Education 0.14(0.09) 0.19(0.11)† −0.001(0.11) 0.07(0.13) −0.39(0.13)* 0.51(0.11)** 0.55(0.12)** 0.38(0.12)* 0.21(0.20) −0.21(0.16)
Intervention 0.15(0.07)* 0.12(0.09) 0.13(0.09) −0.13(0.13) −0.09(0.14) −0.07(0.14) −0.20(0.14) −0.06(0.14) −0.31(0.17)† 0.01(0.14)
Ancestry PC 1 −0.03(0.08) −0.07(0.10) −0.01(0.11) −0.08(0.14) 0.05(0.14) −0.32(0.15)* −0.25(0.16) 0.05(0.13) 0.08(0.23) 0.70(0.11)**
Ancestry PC 2 −0.72(0.13)** −0.75(0.14)** −0.70(0.15)** 0.22(0.17) 0.32(0.18)† −0.27(0.13)* −0.21(0.15) −0.22(0.14) 0.09(0.25) 0.08(0.18)
Sex −0.08(0.09) 0.01(0.10) −0.12(0.11) 0.27(0.13)† −0.05(0.14) 0.06(0.12) 0.15(0.13) 0.17(0.12) 0.23(0.18) 0.23(0.14)†
Par Neg age 2 −0.04(0.05) −0.06(0.07) −0.01(0.06) −0.03(0.15) 0.09(0.12) 0.12(0.19) 0.32(0.18)† 0.38(0.20)† 0.50(0.17)* 0.53(0.15)*
Par Pos age 2 −0.06(0.07) 0.02(0.09) −0.17(0.10)† 0.25(0.13)† −0.23(0.16) 0.12(0.16) 0.27(0.16)† 0.14(0.16) 0.38(0.16)* 0.05(0.15)
Par Neg age 3 −0.05(0.07) 0.08(0.05) 0.03(0.06) 0.37(0.19)* 0.20(0.19) 0.01(0.07) −0.03(0.09) 0.03(0.07) −0.08(0.21) 0.05(0.13)
Par Pos age 3 0.07(0.10) 0.13(0.12) 0.09(0.13) 0.13(0.14) 0.02(0.17) 0.49(0.18)* 0.44(0.19)* 0.60(0.16)** −0.02(0.22) 0.27(0.16)†
Maternal Depression −0.22(0.09)* −0.25(0.11)* −0.30(0.11)* 0.03(0.13) −0.12(0.16) −0.03(0.13) −0.06(0.14) −0.07(0.14) −0.08(0.22) −0.09(0.15)
Internalizing age 2 −0.33(0.10)* −0.18(0.12) −0.33(0.12)* 0.52(0.16)* 0.12(0.16) 0.36(0.17)* 0.34(0.17)† 0.34(0.17)* 0.02(0.26) 0.01(0.16)
Externalizing age 2 −0.23(0.09)* −0.17(0.11) −0.04(0.10) 0.27(0.20) 0.54(0.17)** −0.28(0.16)† −0.19(0.18) −0.25(0.17) 0.14(0.30) 0.04(0.20)
Open in a new tab

Notes. N=515.

**

p<0.001

*

p<0.05

†

p<0.10.

Standardized Coefficients(Standard Errors) presented. CO: Co-occurring. Low: Low Problems. INT: Internalizing ‘only.’ EXT: Externalizing ‘only.’ PC: Principal Component. Par Neg: Parental negative affect. Par Pos: Parental positive affect. BI: Behavioral Inhibition. NA: Negative Affectivity. IC: Inhibitory Control. Sex of child: 0=male, 1=female. Urbanicity: 0=rural, 1=suburban, 2=urban. Intervention: 0=control, 1=intervention.

Negative affectivity

Higher levels of the aggression PRS, but not the internalizing PRS, predicted higher levels of age 3 negative affectivity and explained 0.7% of the variance (adjusted R2=.007). This effect of the aggression PRS on negative affectivity only emerged once parents’ affect was controlled (without adjustment β=.07, p=.07). Negative affectivity at age 2 was not predicted by either PRS.

Behavioral inhibition

Higher internalizing PRSs predicted higher age 2 behavioral inhibition and explained .8% of the variance (adjusted R2=.008). Lower aggression PRS predicted higher age 2 behavioral inhibition and explained 1.7% of the variance (adjusted R2=.017). Behavioral inhibition at age 3 was not predicted by either PRS.

Inhibitory control

Inhibitory control at ages 2 or 3 was not predicted by either PRS.

Temperamental Prediction of Latent Profiles

Negative affectivity

Across both reporters, higher levels of age 3 negative affectivity predicted a higher likelihood of belonging to the co-occurring group relative to all other groups: low problems (primary caregiver: OR=1.72, CI95=[1.23, 2.42]; teacher: OR=1.93, CI95=[1.37, 2.72]), externalizing ‘only’ (primary caregiver: OR=1.92; CI95=[1.26, 2.91]; teacher: OR=2.15, CI95=[1.40, 3.29]), and internalizing ‘only’ (primary caregiver: OR=1.73, CI95=[1.21, 2.46]; teacher: OR=1.72, CI95=[1.08, 2.72]). Age 2 negative affectivity did not differentiate among problem behavior groups for either reporter.

Behavioral inhibition

Higher levels of age 2 behavioral inhibition predicted a higher likelihood of membership in the co-occurring relative to externalizing ‘only’ group, as reported by primary caregivers (significantly; OR=1.73, CI95=[1.07, 2.56]) and by teachers (marginally significantly; OR=1.51, CI95=[0.75, 2.40]). For primary caregivers only, higher behavioral inhibition predicted membership in the co-occurring relative to low problems group (OR=1.66, CI95=[1.04, 2.63]). For teacher reports only, higher levels of age 2 behavioral inhibition predicted membership in the internalizing ‘only’ relative to low problems group (OR=1.40, CI95=[1.04, 1.90]). Age 3 behavioral inhibition did not differentiate any problem behavior groups.

Inhibitory control

Neither age 2- or age 3-measured inhibitory control differentiated any of the problem behavior groups.

PRS Prediction of Latent Profiles

Higher levels of the internalizing PRS predicted membership in the primary caregiver-reported co-occurring group relative to the internalizing ‘only’ (OR=1.84, CI95=[0.34, 0.77]) and externalizing ‘only’ groups (OR=1.94, CI95=[0.31, 0.85]). Higher internalizing PRS predicted membership in the low problems group relative to internalizing ‘only’ group as reported by both primary caregivers (OR=1.42, CI95=[1.08, 1.87]) and teachers (marginally significantly; OR=1.32, CI95=[0.95, 1.81]), and membership in the low problems group relative to the externalizing ‘only’ group for primary caregivers only (OR=1.50, CI95=[1.09, 2.04]).

Mediation

According to the joint significance test, higher levels of the aggression PRS indirectly predicted primary caregiver- and teacher-reported co-occurring problems (relative to externalizing ‘only,’ internalizing ‘only,’ and low problems) through age 3 negative affectivity. Higher levels of the internalizing PRS and lower levels of the aggression PRS each indirectly predicted primary caregiver-reported co-occurring problems (relative to externalizing ‘only’ and low problems) through age 2 behavioral inhibition. Higher levels of the internalizing PRS and lower levels of the aggression PRS also indirectly predicted teacher-reported internalizing problems ‘only’ (relative to low problems) through age 2 behavioral inhibition.

Secondary Analyses: Bifactor Model of Psychopathology

See supplementary Table 4 for full results.

General factor

Higher levels of age 3 negative affectivity in 3-year-olds marginally significantly predicted higher levels of the general factor. Lower levels of inhibitory control in 3-year-olds significantly predicted higher levels of the general factor. No other temperament dimensions predicted the general factor.

Externalizing and internalizing factors

Higher levels of behavioral inhibition in 2-year olds marginally predicted higher levels of externalizing and internalizing factors. Higher levels of age-3 inhibitory control significantly predicted higher externalizing and internalizing factors.

Discussion

Distinct pathways to co-occurring internalizing and externalizing problems in adolescence are poorly understood, hindering the development of personalized approaches to prevention and intervention. By capitalizing on multiple levels of analysis, including genetics, observed behaviors, and questionnaires from multiple reporters, we found evidence for two pathways to co-occurring problems in adolescence stemming from unique early childhood genetic and temperamental factors. In discussing our findings, we focus on effects that showed evidence of replication (significant or marginally significant) with a second reporter.

Negative Affect

One mediational pathway to LPA-defined co-occurring problems (relative to internalizing ‘only,’ externalizing ‘only,’ and low problems) stemmed from higher levels of age 3 negative affect, which was predicted by higher levels of the aggression PRS. This particularly robust effect replicated with both reporters of problem behaviors and was not confounded by reporter bias, as negative affect was observed. Bifactor analyses were consistent, albeit only marginally significant, showing that children with higher levels of genetically-influenced negative affectivity were somewhat more likely to exhibit greater general psychopathology. Results are in line with prior findings showing that what underlies the shared genetic basis of internalizing and externalizing problems may be, in part, negative affectivity (Hink et al., 2013; Mikolajewski et al., 2013; Tackett et al., 2011, 2013). Moreover, this genetically-influenced, shared risk factor can be detected as early as age 3, a novel finding herein. Notably, externalizing-related aspects of negative affectivity, such as antagonism, are more observable than is internalized distress and may be overweighted in measures of negative affectivity (e.g., Tackett et al., 2012). This overweighting may explain why only the aggression PRS was related to negative affect and may further suggest that genetically-influenced externalizing-related negative affect creates generalized risk for internalizing and externalizing problems.

Behavioral Inhibition

A different pathway to LPA-defined co-occurring problems also was evident, characterized by higher levels of behavioral inhibition at age 2, which was predicted by lower levels of the aggression PRS and higher levels of the internalizing PRS. Specifically, behavioral inhibition differentiated adolescents in the co-occurring vs. externalizing ‘only’ groups (significantly for primary caregivers and marginally for teachers). For one reporter only, behavioral inhibition was higher for adolescents in the co-occurring group (primary caregiver) and the internalizing ‘only’ group (teachers) relative to the low problems group. Our confidence in the contribution of behavioral inhibition to co-occurring problems was strengthened by the lack of confounding by reporter bias (i.e., behavioral inhibition was observed). Somewhat consistently, bifactor analyses demonstrated that children showing greater genetically-influenced behavioral inhibition were marginally significantly more likely to exhibit higher levels of both specific externalizing and internalizing latent factors. Taken together, results tentatively suggest that one reason behavioral inhibition predicts co-occurring problems is because it generates features of internalizing or externalizing problems that are unique from one another and from generalized risk for psychopathology. Although behavioral inhibition is most often associated with later internalizing problems (De Pauw & Mervielde, 2010), perhaps early behavioral inhibition is a marker of later aggressive or disruptive behaviors that result from the channeling of internalized distress outwards (Zhan et al., 2015). This mechanistic process to co-occurrence may also be more commonly observed at home than at school, based on more robust effects of behavioral inhibition on primary caregiver-reported outcomes.

Integrating genetic indices

Importantly, there were opposing effects of the aggression PRS on behavioral inhibition (negative) and negative affectivity (positive), yet each temperamental dimension increased risk for co-occurring problems. This pattern of findings suggests that these dimensions of temperament may mark two separate genetic pathways through which adolescents develop co-occurring problems. Without indices of genetic risk, we may have concluded that children with co-occurring problems simply showed elevations in both behavioral inhibition and negative affectivity. Including measures of genetic risk shed light on the heterogeneous etiologic origins of co-occurring problems that would otherwise appear equivalent. Moreover, the divergent associations of behavioral inhibition and negative affectivity on the factors of the bifactor model further reinforced the mechanistic distinctiveness of these pathways to co-occurring problems. Evidence for equifinality to co-occurring problems further implies that intervention strategies for co-occurring problems may need to be tailored depending on the pathways children follow to this severe outcome. An important avenue for future research will be to understand whether indirect effects reflect distinctive subgroups of children and, subsequently, to identify whether unique treatment strategies are needed for each.

Inhibitory Control

Inhibitory control did not predict latent profiles of adolescents’ problem behaviors. However, in bifactor analyses, lower levels of age 3 inhibitory control predicted greater general psychopathology, whereas higher levels predicted greater specific internalizing and externalizing problems. Perhaps the co-occurring latent profile did not show associations with inhibitory control because it may be comprised of a mixture of those with risk for either generalized psychopathology (predicted by low inhibitory control) or for both specific internalizing and externalizing (predicted by high inhibitory control). Nonetheless, bifactor model results are consistent with prior research showing that lower effortful control, of which inhibitory control is one facet, predicts generalized risk for psychopathology (Shields et al., 2019). Offering one explanation for its effect on the specific internalizing factor, extremely high inhibitory control can, in some instances, capture overcontrolled behavior (Murray & Kochanska, 2002). Although high inhibitory control was not expected to predict the specific externalizing factor, perhaps similar to associations with behavioral inhibition, high inhibitory control reflects an internalizing pathway to co-occurring problems.

Age-Specific Effects

PRSs only predicted behavioral inhibition at age 2 and negative affectivity at age 3. Interestingly, the co-occurring group, derived from LPAs, was also only related to behavioral inhibition at age 2 and negative affectivity at age 3, suggesting more precise measurement of temperament in our study at these ages. Age-specific findings may not be surprising, as consistent with prior research (e.g., Majdandžić & Boom, 2007), our observed dimensions of temperament showed low-to-moderate stability over time (see Table 1). Indeed, observations of behavioral inhibition and negative affectivity by independent raters at each time point would not be expected to be as stable as if they were reported by the same informant over time. Nonetheless, replication is needed to fully alleviate concerns about age-specific effects.

Direct Genetic Effects

The aggression PRS only predicted age 3 negative affectivity once controlling for parents’ observed affect during the same tasks. Genetic effects have been shown to strengthen after controlling for environmental variation (Leve et al., 1998), suggesting the importance of doing so in future studies. Surprisingly, higher internalizing PRSs predicted the low problems relative to the internalizing ‘only’ group as reported by primary caregivers (significant) and teachers (marginal). Yet, higher internalizing PRSs also predicted risk for co-occurring problems indirectly through behavioral inhibition. Perhaps after accounting for behavioral inhibition, the internalizing PRS reflects protective internalizing-related characteristics. For example, fearful arousal predicted a greater internalized conscience for children whose parents de-emphasized the use of power in caregiving (Kochanska, 1991b). More research is needed to confirm these unexpected findings.

Strengths and Limitations

Our study should be interpreted in the context of several limitations. We did not test interactions of PRSs or temperament with other variables, such as the sex of the child, or PRS-by-PRS interactions, because the model was already complex with 14–22 predictors. However, as intervention group status is an important environmental variable that could change how age 3 temperament predicts adolescents’ latent profiles, we conducted post-hoc temperament-by-intervention interactions and found none. PRSs, derived from GWASs of European children, may not capture genetic risk for non-European participants (although ancestry was controlled). PRSs showed weak associations with temperament and problem behavior and did not inform biological mechanisms, consistent with prior work in this sample with PRSs (e.g., Elam et al., 2018). In the future, PRSs constructed from independent GWASs that are sufficiently large to detect smaller effect sizes than the currently used GWASs and/or that take into consideration functional genomic data (Kichaev et al., 2019) may explain greater proportions of variance. No additional temperament dimensions were assessed in this study, precluding their use (e.g., positive affectivity). Latent profile groups simply reflect patterns of symptomatology and may be difficult to replicate; thus, groups should not be reified and caution is needed in interpreting results without replication. It is unknown whether results would generalize to children living in lower-risk communities, where better childcare, neighborhood, and school quality may buffer effects of genetic and temperamental risk factors. Finally, we were unable to include childreported problem behavior, as youth only reported a small number of different problems.

This study has strengths that reinforce confidence in findings. Adolescent problem behaviors were assessed using multiple informants and characterized using empirically-based latent variable methods. We integrated measures of genetic risk and observers’ ratings of children’s temperament into longitudinal models, allowing us to more fully characterize youths at-risk for co-occurring problems. Our multi-method approach alleviated concerns of reporter bias, increasing confidence in findings even when replication across reporters was not consistent.

Conclusions

This study provided evidence for two pathways to co-occurring problem behaviors that warrant additional follow-up: one stemming from genetically-influenced negative affectivity and the other from genetically-influenced behavioral inhibition. Results call for future research to confirm these pathways and to test environmental moderators to determine the best forms of intervention for each. Findings point to the exciting possibility that, with advances in our knowledge of the human genome, precision-medicine based prevention programs may be feasible in the future.

Supplementary Material

Supplemental Material

Acknowledgements

This research was principally supported by grants 023245 and 036832 from NIDA (Shaw, Dishion, Wilson). Additional support was provided by 007453 from NIAAA (Richardson, Molina).

This research was approved by the institutional review boards at the University of Pittsburgh (PRO13070142), University of Virginia (HSR#17795), and Arizona State University (IRB#00000278).

Footnotes

1

An aggression PRS, rather than a broader externalizing PRS, was examined because no sufficiently large GWAS of externalizing problems has been conducted during early childhood, the period of interest. However, aggression is a good proxy for externalizing problems as it is more prevalent than rule-breaking behavior in childhood and is highly genetically-correlated with other forms of externalizing behavior (Niv et al., 2013).

References

  1. Achenbach TM, & Rescorla LA (2000). Manual for the Child Behavior Checklist/1½−5 and 2000 Profile. Burlington, VT: University of Vermont, Research Center for Children, Youth, & Families. [Google Scholar]
  2. Achenbach TM, & Rescorla LA (2001). Manual for the ASEBA school-age forms & profiles: Child behavior checklist for ages 6–18, teacher’s report form, youth self-report: an integrated system of multi-informant assessment. University of Vermont, research center for children youth & families. [Google Scholar]
  3. Benke KS, Nivard MG, Velders FP, Walters RK, Pappa I, Scheet PA, Xiao X, Ehli EA, Palmer LJ, & Whitehouse AJ (2014). A genome-wide association meta-analysis of preschool internalizing problems. Journal of the American Academy of Child & Adolescent Psychiatry, 53(6), 667–676. [DOI] [PubMed] [Google Scholar]
  4. Capaldi DM (1992). Co-occurrence of conduct problems and depressive symptoms in early adolescent boys: II. A 2-year follow-up at Grade 8. Development and Psychopathology, 4(1), 125–144. [DOI] [PubMed] [Google Scholar]
  5. Caron C, & Rutter M (1991). Comorbidity in child psychopathology: Concepts, issues and research strategies. Journal of Child Psychology and Psychiatry, 32(7), 1063–1080. [DOI] [PubMed] [Google Scholar]
  6. Caspi A, Moffitt TE, Newman DL, & Silva PA (1996). Behavioral observations at age 3 years predict adult psychiatric disorders: Longitudinal evidence from a birth cohort. Archives of General Psychiatry, 53(11), 1033–1039. [DOI] [PubMed] [Google Scholar]
  7. Chen FF, West S, & Sousa K (2006). A Comparison of Bifactor and Second-Order Models of Quality of Life. Multivariate Behavioral Research, 41(2), 189–225. [DOI] [PubMed] [Google Scholar]
  8. Cosgrove VE, Rhee SH, Gelhorn HL, Boeldt D, Corley RC, Ehringer MA, Young SE, & Hewitt JK (2011). Structure and etiology of co-occurring internalizing and externalizing disorders in adolescents. Journal of Abnormal Child Psychology, 39(1), 109–123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Dahl RE, Allen NB, Wilbrecht L, & Suleiman AB (2018). Importance of investing in adolescence from a developmental science perspective. Nature, 554, 441–450. [DOI] [PubMed] [Google Scholar]
  10. De Pauw SS, & Mervielde I (2010). Temperament, personality and developmental psychopathology: A review based on the conceptual dimensions underlying childhood traits. Child Psychiatry & Human Development, 41(3), 313–329. [DOI] [PubMed] [Google Scholar]
  11. Diamantopoulou S, Verhulst FC, & Van Der Ende J (2011). Gender differences in the development and adult outcome of co-occurring depression and delinquency in adolescence. Journal of Abnormal Psychology, 120(3), 644. [DOI] [PubMed] [Google Scholar]
  12. Dishion TJ, Shaw D, Connell A, Gardner F, Weaver C, & Wilson M (2008). The family check-up with high-risk indigent families: Preventing problem behavior by increasing parents’ positive behavior support in early childhood. Child Development, 79(5), 1395–1414. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Edwards RC, & Hans SL (2015). Infant risk factors associated with internalizing, externalizing, and co-occurring behavior problems in young children. Developmental Psychology, 51(4), 489. [DOI] [PubMed] [Google Scholar]
  14. Elam KK, Chassin L, & Pandika D (2018). Polygenic risk, family cohesion, and adolescent aggression in Mexican American and European American families: Developmental pathways to alcohol use. Development and Psychopathology, 30(5), 1715–1728. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Euesden J, Lewis CM, & O’Reilly PF (2015). PRSice: Polygenic risk score software. Bioinforma Oxf Engl 31: 1466–1468. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Evans DM, Brion MJA, Paternoster L, Kemp JP, McMahon G, Munafò M, Whitfield JB, Medland SE, Montgomery GW, & Timpson NJ (2013). Mining the human phenome using allelic scores that index biological intermediates. PLoS Genetics, 9(10), e1003919. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fanti KA, & Henrich CC (2010). Trajectories of pure and co-occurring internalizing and externalizing problems from age 2 to age 12: Findings from the National Institute of Child Health and Human Development Study of Early Child Care. Developmental Psychology, 46(5), 1159. [DOI] [PubMed] [Google Scholar]
  18. Gillespie NA, Cloninger CR, Heath AC, & Martin NG (2003). The genetic and environmental relationship between Cloninger’s dimensions of temperament and character. Personality and Individual Differences, 35(8), 1931–1946. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Goldsmith HH, Buss KA, & Lemery KS (1997). Toddler and childhood temperament: Expanded content, stronger genetic evidence, new evidence for the importance of environment. Developmental Psychology, 33(6), 891. [DOI] [PubMed] [Google Scholar]
  20. Grotzinger AD, Cheung AK, Patterson MW, Harden KP, & Tucker-Drob EM (2019). Genetic and Environmental Links Between General Factors of Psychopathology and Cognitive Ability in Early Childhood. Clinical Psychological Science, 7(3), 430–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hankin BL, Davis EP, Snyder H, Young JF, Glynn LM, & Sandman CA (2017). Temperament factors and dimensional, latent bifactor models of child psychopathology: Transdiagnostic and specific associations in two youth samples. Psychiatry Research, 252, 139–146. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Hannigan LJ, Pingault JB, Krapohl E, McAdams TA, Rijsdijk FV, & Eley TC (2018). Genetics of co-developing conduct and emotional problems during childhood and adolescence. Nature Human Behaviour, 2(7), 514. [DOI] [PubMed] [Google Scholar]
  23. Hink LK, Rhee SH, Corley RP, Cosgrove VE, Hewitt JK, Schulz-Heik RJ, Lahey BB, & Waldman ID (2013). Personality dimensions as common and broadband-specific features for internalizing and externalizing disorders. Journal of Abnormal Child Psychology, 41(6), 939–957. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Keiley MK, Lofthouse N, Bates JE, Dodge KA, & Pettit GS (2003). Differential risks of covarying and pure components in mother and teacher reports of externalizing and internalizing behavior across ages 5 to 14. Journal of Abnormal Child Psychology, 31(3), 267–283. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kichaev G, Bhatia G, Loh PR, Gazal S, Burch K, Freund MK, Schoech A, Pasaniuc B, & Price AL (2019). Leveraging Polygenic Functional Enrichment to Improve GWAS Power. The American Journal of Human Genetics, 104(1), 65–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kochanska G (1991a). Patterns of Inhibition to the Unfamiliar in Children of Normal and Affectively Ill Mothers. Child Development, 62(2), 250–263. [DOI] [PubMed] [Google Scholar]
  27. Kochanska G (1991b). Socialization and Temperament in the Development of Guilt and Conscience. Child Development, 62(6), 1379–1392. [PubMed] [Google Scholar]
  28. Kochanska G, Murray KT, & Harlan ET (2000). Effortful control in early childhood: Continuity and change, antecedents, and implications for social development. Developmental Psychology, 36(2), 220–232. 10.1037/0012-1649.36.2.220 [DOI] [PubMed] [Google Scholar]
  29. Lahey BB, Van Hulle CA, Singh AL, Waldman ID, & Rathouz PJ (2011). Higher-order genetic and environmental structure of prevalent forms of child and adolescent psychopathology. Archives of General Psychiatry, 68(2), 181–189. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Lemery KS, Goldsmith HH, Klinnert MD, & Mrazek DA (1999). Developmental models of infant and childhood temperament. Developmental Psychology, 35(1), 189. [DOI] [PubMed] [Google Scholar]
  31. Lemery-Chalfant K, Clifford S, Dishion TJ, Shaw DS, & Wilson MN (2018). Genetic moderation of the effects of the Family Check-Up intervention on children’s internalizing symptoms: A longitudinal study with a racially/ethnically diverse sample. Development and Psychopathology, 30(5), 1729–1747. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lemery-Chalfant K, Doelger L, & Goldsmith HH (2008). Genetic relations between effortful and attentional control and symptoms of psychopathology in middle childhood. Infant and Child Development, 17(4), 365–385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Leve LD, Winebarger AA, Fagot BI, Reid JB, & Goldsmith HH (1998). Environmental and genetic variance in children’s observed and reported maladaptive behavior. Child Development, 69(5), 1286–1298. [PubMed] [Google Scholar]
  34. Loeber R, & Keenan K (1994). Interaction between conduct disorder and its comorbid conditions: Effects of age and gender. Clinical Psychology Review, 14(6), 497–523. [Google Scholar]
  35. MacKinnon DP, Lockwood CM, Hoffman JM, West SG, & Sheets V (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Majdandžić M, & Boom DCVD (2007). Multimethod Longitudinal Assessment of Temperament in Early Childhood. Journal of Personality, 75(1), 121–168. [DOI] [PubMed] [Google Scholar]
  37. Mikolajewski AJ, Allan NP, Hart SA, Lonigan CJ, & Taylor J (2013). Negative Affect Shares Genetic and Environmental Influences with Symptoms of Childhood Internalizing and Externalizing Disorders. Journal of Abnormal Child Psychology, 41(3), 411–423. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Moilanen KL, Shaw DS, & Maxwell KL (2010). Developmental cascades: Externalizing, internalizing, and academic competence from middle childhood to early adolescence. Development and Psychopathology, 22(3), 635–653. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Murray KT, & Kochanska G (2002). Effortful Control: Factor Structure and Relation to Externalizing and Internalizing Behaviors. Journal of Abnormal Child Psychology, 30, 503–514. [DOI] [PubMed] [Google Scholar]
  40. Muthén, & Muthén. (1998). Mplus User’s Guide (Eigth). Muthén & Muthén. [Google Scholar]
  41. Neumann A, Pappa I, Lahey BB, Verhulst FC, Medina-Gomez C, Jaddoe VW, Bakermans-Kranenburg MJ, Moffitt TE, van IJzendoorn MH, & Tiemeier H (2016). Single Nucleotide Polymorphism Heritability of a General Psychopathology Factor in Children. Journal of the American Academy of Child & Adolescent Psychiatry, 55(12), 1038–1045.e4. [DOI] [PubMed] [Google Scholar]
  42. Niv S, Tuvblad C, Raine A, & Baker LA (2013). Aggression and rule-breaking: Heritability and stability of antisocial behavior problems in childhood and adolescence. Journal of Criminal Justice, 41(5), 285–291. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Nylund KL, Asparouhov T, & Muthén BO (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569. [Google Scholar]
  44. O’Connor TG, McGuire S, Reiss D, Hetherington EM, & Plomin R (1998). Co-occurrence of depressive symptoms and antisocial behavior in adolescence: A common genetic liability. Journal of Abnormal Psychology, 107(1), 27. [DOI] [PubMed] [Google Scholar]
  45. Oland AA, & Shaw DS (2005). Pure versus co-occurring externalizing and internalizing symptoms in children: The potential role of socio-developmental milestones. Clinical Child and Family Psychology Review, 8(4), 247–270. [DOI] [PubMed] [Google Scholar]
  46. Olino TM, Dougherty LR, Bufferd SJ, Carlson GA, & Klein DN (2014). Testing Models of Psychopathology in Preschool-aged Children Using a Structured Interviewbased Assessment. Journal of Abnormal Child Psychology, 42(7), 1201–1211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Pappa I, St Pourcain B, Benke K, Cavadino A, Hakulinen C, Nivard MG, Nolte IM, Tiesler CM, Bakermans-Kranenburg MJ, & Davies GE (2016). A genome-wide approach to children’s aggressive behavior: The EAGLE consortium. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 171(5), 562–572. [DOI] [PubMed] [Google Scholar]
  48. Peterson J, Winter C, Jabson J, & Dishion TJ (2008). Relationship affect coding system. Unpublished Coding Manual, University of Oregon, Child and Family Center, Eugene [Google Scholar]
  49. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P, De Bakker PI., & Daly MJ (2007). PLINK: A tool set for whole-genome association and population-based linkage analyses. The American Journal of Human Genetics, 81(3), 559–575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Radloff LS (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1(3), 385–401. [Google Scholar]
  51. Rothbart MK, Ahadi SA, Hershey KL, & Fisher P (2001). Investigations of temperament at three to seven years: The Children’s Behavior Questionnaire. Child Development, 72(5), 1394–1408. [DOI] [PubMed] [Google Scholar]
  52. Sanson A, Hemphill SA, & Smart D (2004). Connections between Temperament and Social Development: A Review. Social Development, 13(1), 142–170. [Google Scholar]
  53. Shields AN, Reardon KW, Brandes CM, & Tackett J L. (2019). The p factor in children: Relationships with executive functions and effortful control. Journal of Research in Personality, 82, 103853. [Google Scholar]
  54. Tackett JL, Lahey BB, Hulle CV, Waldman I, Krueger RF, & Rathouz PJ (2013). Common Genetic Influences on Negative Emotionality and a General Psychopathology Factor in Childhood and Adolescence. Journal of Abnormal Psychology, 122(4), 1142–1153. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Tackett JL, Slobodskaya HR, Mar RA, Deal J, Halverson CF, Baker SR, Pavlopoulos V, & Besevegis E (2012). The Hierarchical Structure of Childhood Personality in Five Countries: Continuity From Early Childhood to Early Adolescence. Journal of Personality, 80(4), 847–879. [DOI] [PubMed] [Google Scholar]
  56. Tackett JL, Waldman ID, Van Hulle CA, & Lahey BB (2011). Shared Genetic Influences on Negative Emotionality and Major Depression/Conduct Disorder Comorbidity. Journal of the American Academy of Child and Adolescent Psychiatry, 50(8), 818–827. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Wang FL, Eisenberg N, Valiente C, & Spinrad TL (2016). Role of temperament in early adolescent pure and co-occurring internalizing and externalizing problems using a bifactor model: Moderation by parenting and gender. Development and Psychopathology, 28(4pt2), 1487–1504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Wang FL, Feldman JS, Lemery-Chalfant K, Wilson MN, & Shaw DS (2019). Family-based prevention of adolescents’ co-occurring internalizing/externalizing problems through early childhood parent factors. Journal of Consulting and Clinical Psychology, 87(11), 1056–1067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Zhan J, Ren J, Fan J, & Luo J (2015). Distinctive effects of fear and sadness induction on anger and aggressive behavior. Frontiers in Psychology, 6, 725. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Material
NIHMS1590071-supplement-Supplemental_Material.pdf (357.7KB, pdf)

RESOURCES