A Population-Based Twin Study of Childhood Irritability and Internalizing Syndromes

Abstract

Objective:

Childhood irritability exhibits significant theoretical and empirical associations with depression and anxiety syndromes. The current study used the twin design to parse genetic and environmental contributions to these relationships.

Method:

Children aged 9–14 from 374 twin pairs were assessed for irritability and symptoms of depression, generalized anxiety, panic, social phobia, and separation anxiety using dimensional self-report instruments. Multivariate structural equation modeling decomposed the correlations between these syndromes into genetic and environmental components to examine shared and specific risk domains.

Results:

Irritability had significant associations with each internalizing symptom domain. Genetic contributions to irritability are moderately correlated with genetic risk for symptoms of depression, generalized anxiety, and separation anxiety with weaker overlap with the other anxiety syndromes. Familial and specific environmental risk factors explained covariation among syndromes and indicated potential syndrome-specific risk.

Conclusions:

There is substantial overlap among the genetic and environmental factors that influence individual differences in irritability and those that increase liability for depression and anxiety symptoms in children. These findings deepen the current understanding of childhood internalizing risk factors and provide important implications for syndrome prediction and susceptibility gene discovery efforts.

Internalizing disorders (IDs), comprised of depressive and anxiety disorders, create considerable burden (Murray & Lopez, 1997), partly because they frequently arise in childhood and persist throughout adulthood (Maser & Cloninger, 1990). Both adult (Hettema, Neale, Myers, Prescott, & Kendler, 2006) and child (Waszczuk, Zavos, Gregory, & Eley, 2014) twin studies denote common genetic and environmental contributions to depressive and anxiety disorders, which may explain the comorbidity observed in childhood and across the life course (Middeldorp, Cath, Van Dyck, & Boomsma, 2005). Additionally, multi- and equifinality in the development of IDs (Cicchetti & Rogosch, 2002) emphasize the importance of clarifying similarities and distinctions in their etiology. A deeper etiologic understanding of children’s irritability proneness could provide insights into the pathways underlying a wide range of internalizing psychopathology.

Childhood irritability, the proneness to anger relative to peers (Brotman, Kircanski, Stringaris, Pine, & Leibenluft, 2017), has been identified as an important correlate and risk factor of IDs (see Vidal-Ribas, Brotman, Valdivieso, Leibenluft, & Stringaris, 2016 for review). Furthermore, irritability is often a clinical symptom seen in major depression and generalized anxiety, particularly among children (American Psychiatric Association, 2013). Across development, irritability is associated with elevated emotional and behavioral responding to threat or frustrative non-reward (see Brotman et al., 2017 for review), the latter of which has been proposed as a transdiagnostic risk factor in the negative valence domain of the National Institute of Mental Health Research Domain Criteria (Insel et al., 2010). Given its cross-cutting characteristics as both a normative affective response to frustration (Wakschlag et al., 2015) and a precursor to a broad range of childhood syndromes, irritability might provide a common marker of, or pathway for, various domains of psychopathology. While the association of irritability with externalizing syndromes is well-documented (e.g., Wakschlag et al., 2015), less is known about the mechanisms by which early irritability is associated with mood and anxiety disorders.

Three child and adolescent twin studies have examined irritability, reporting it to be moderately heritable (Mikolajewski, Taylor, & Iacono, 2017; Roberson-Nay et al., 2015; Stringaris, Zavos, Leibenluft, Maughan, & Eley, 2012). An extension of Roberson-Nay et al. (2015) reports that common genetic and environmental factors may influence irritability and internalizing symptoms from middle childhood to young adulthood (Savage et al., 2015). A longitudinal study suggests that irritability may predict new-onset depression in adolescents at increased familial risk (Rice et al., 2017). Mikolajewski, Taylor, and Iacono (2017) provide data among 11-year old children indicating that irritability may be associated with overall internalizing disorder risk at age 17 years, including increased symptoms of major depressive disorder, specific phobia, and social phobia assessed using a structured diagnostic interview.

However, prior studies are limited. Primarily, prior research has examined only the association of irritability with general internalizing disorder risk. For example, Savage and colleagues (2015) demonstrate an association of irritability with a composite of depression and anxiety severity drawn from the Child Behavior Checklist. There is a critical gap in the literature to clarify the contribution of common genetic or environmental factors to the association of irritability with specific internalizing syndromes. To our knowledge, no studies have attempted to dissect the relative genetic contributions of irritability to anxiety versus depressive outcomes or contributions of irritability to the range of anxiety disorder symptoms manifested in the general population. Secondly, further research is required to examine the distribution of irritability and internalizing disorders as they manifest in the community; symptom counts derived from clinical interview may not adequately reflect the distribution of ID symptoms in the population (e.g., Hudziak, Achenbach, Althoff, & Pine, 2007). Finally, prior research in this area (see above) extracted an assessment of irritability from broad assessments of other psychopathology. For example, Roberson-Nay and colleagues (Roberson-Nay et al., 2015; Savage et al., 2015) created a measure of irritability from scales of the Child Behavior Checklist (Achenbach, 1991), while Mikolajewski, Taylor, and Iacono (2017) derived irritability from oppositional defiant disorder symptoms in a diagnostic interview. This procedure risks conflating irritability with related syndromes due to shared items (e.g., sadness) or highly correlated items initially generated to assess the same trait (e.g., symptoms of oppositional defiant disorder).

Further research is needed to understand the current conceptualization of irritability, which has informed neuroscience and developmental psychopathology (see Vidal-Ribas et al., 2016 for review). The present study used a specially designed scale to measure the construct of irritability, the Affective Reactivity Index (ARI; Stringaris, Goodman, et al., 2012). While our group used the present sample to demonstrate that irritability is phenotypically correlated with other aspects of dysphoric affect (i.e., depression, anxiety, and neuroticism) (Lee et al., 2017), no empirical evidence has yet been presented to dissect differential genetic and environmental relationships between irritability and depression versus anxiety or specific childhood anxiety domains. Identifying common and differential genetic and environmental overlap of irritability with depression and anxiety syndromes may clarify the role of irritability to explain the observed high comorbidity between mood and anxiety disorders. This study addresses these knowledge gaps by attempting to answer the following key questions:

• (1)What is the heritability of childhood irritability as assessed by the ARI?
• (2)Is irritability associated with all childhood internalizing disorders i.e., depression and each anxiety syndrome?
• (3)What role do the genetic and environmental factors underlying childhood inter-individual differences in irritability play in risk for depression versus the various anxiety syndromes?
• (4)Do irritability-related etiological factors account for the observed comorbidity and overlap in risk between depression and anxiety syndromes?

Given (i) the symptomatic role of irritability in depression and generalized anxiety and (ii) previously established strong genetic overlap between depression and generalized anxiety in adults (e.g., Hettema et al., 2006), we hypothesized that irritability would exhibit a particularly strong relationship with symptoms of depression and generalized anxiety in children. In supplemental analyses, we then examine the contribution of irritability to externalizing disorders and investigate whether, when accounting for a mutual influence of irritability, genetic and environmental influences to internalizing psychopathology are correlated with genetic and environmental influences to externalizing psychopathology.

Methods

Participants

The present study used data on 748 twin children (131 monozygotic and 243 dizygotic pairs) aged 9 to 14 from the Virginia Commonwealth University Juvenile Anxiety Study (VCU-JAS; [REMOVED FOR MASKED REVIEW]). This genetic epidemiological sample includes Caucasian children recruited through the Mid-Atlantic Twin Registry (Lilley & Silberg, 2013). Only Caucasian families were included to minimize phenotypic and genetic heterogeneity for the overall study. Exclusion criteria included intellectual disability or autism spectrum disorder, past or current psychotic episode, current use of anxiolytic or antidepressant medication, or any medical condition which might have adversely impacted participants’ safety or ability to complete the study. Clinical assessment by a study clinician of parent-report on the K-SADS (Kaufman et al., 1997) indicated that the sample is representative of IDs in this age group (e.g., Fergusson, Horwood, & Lynskey, 1993): any depressive disorder (2.2%), any anxiety disorder (27.4%). Participants were recruited at two sites: The VCU Virginia Institute for Psychiatric and Behavioral Genetics (Richmond, VA) and the National Institute for Mental Health Intramural Program (Bethesda, MD). The study protocol was approved by human subjects review boards at both institutions. Consent was obtained from participants’ parents or legal guardians; assent was obtained from participants.

Procedure

Participants completed dimensional assessments of irritability, anxiety, and depression during a laboratory-based session. Parents completed the same dimensional assessments of irritability, anxiety, depression plus externalizing syndromes twice: about each child separately. Participants and parents completed additional instruments and tasks to address additional aims of the VCU-JAS; see [REMOVED FOR MASKED REVIEW] for a full description of the study protocol.

Measures

Irritability.

Participant irritability was assessed via child- and parent-report on the Affective Reactivity Index (ARI), which was developed to assess pediatric irritability (Stringaris, Goodman, et al., 2012). Children completed six questions to assess behavior and feelings of irritability and anger over the past 6 months. Reliability and validity of the ARI among child samples is supported in prior research (Stringaris, Goodman, et al., 2012) and in the current sample ([REMOVED FOR MASKED REVIEW]).

Depression.

Participant depressive symptoms were assessed via child- and parent-report on the Short Mood and Feelings Questionnaire (SMFQ), which is a subset of items from the larger Mood and Feelings Questionnaire (Angold et al., 1995). The SMFQ was modified to assess depressive symptoms over the past 3 months for consistency with the assessment of anxiety symptoms. Reliability and validity of the SMFQ among child samples is supported in prior research (Klein, Dougherty, & Olino, 2005) and in the present study ([REMOVED FOR MASKED REVIEW]).

Anxiety.

Participant anxiety symptoms over the past 3 months were assessed via child- and parent-report on the Screen for Child Anxiety Related Emotional Disorders (SCARED), which contains 41 items that form five correlated subscales assessing symptoms of panic/somatization, generalized anxiety, separation anxiety, social phobia, and school avoidance (Birmaher et al., 1997). Reliability and validity of the SCARED among child samples is supported in prior research (Birmaher et al., 1997) and in the present study ([REMOVED FOR MASKED REVIEW]). Similar to prior research (Lau, Gregory, Goldwin, Pine, & Eley, 2007), only the first four symptom subscales were included in subsequent analyses.

Externalizing Syndromes.

Parents rated externalizing symptoms of each child on the Child Behavioral Checklist (Achenbach, 1991). Severity of attention deficit hyperactivity disorder (ADHD), oppositional defiant disorder, and conduct disorder were extracted based on the DSM-oriented scales. Reliability and validity of the Child Behavior Checklist is supported in prior research (e.g., Ebesutani et al., 2010). In the present sample, inter-item reliability is high for ADHD (based on 7 items; α = 0.89), oppositional defiant disorder (based on 5 items; α = 0.86), and conduct disorder (based on 17 items; α = 0.89).

Data Analysis

The biometrical model, which leverages differential genetic relatedness between monozygotic and dizygotic twins, was used to estimate the influence of latent additive genetic (A), family environmental (C), and unique environmental (E) factors on irritability as well as their influence on the correlation of irritability with each internalizing symptom domain (Neale & Cardon, 1992). An initial, univariate model decomposed variance in irritability to that explained by A, C, and E for i) child-report irritability and ii) a latent irritability phenotype indexed by child- and parent-report. Next, phenotypic correlations of irritability with depression and anxiety disorder symptoms were estimated within a biometrical genetic model to account for nonindependence of participants due to nesting within family including greater nonindependence among monozygotic twins.

While multiple informants are typically preferred, parent-report poses a unique challenge to research that relies on the analysis of twins. Specifically, there are known tendencies for traits of any rater (e.g., a parent) to influence their assessment of each target (e.g., each child; Kenny, Kashy, & Cook, 2006). For example, parental depression symptoms may influence how a given parent rates both children, which would inflate the resulting correlation between twins. Additionally, as described below, twin zygosity was assessed based on parent-report of physical similarities between children. This method shows high agreement with zygosity assessed from DNA (Jackson, Snieder, Davis, & Treiber, 2001; Kasriel & Eaves, 1976) and relies on questions such as rating how frequently close family (e.g., parents and other siblings) have difficulty “telling your twins apart.” Parents who identify that their children are monozygotic have indicated that they, or other close family, have difficulty distinguishing between children, which may lead them to rate monozygotic twins as more similar than a parent whose children are more easily distinguished (i.e., dizygotic). Therefore, the similarity between monozygotic twins will lead to an inflated correlation as compared to dizygotic twins, which will inflate the estimate of genetic influence to each trait and to the correlation between traits. To avoid the potential for inflated correlations and genetic estimates, the present study used only child-report where possible.

Finally, multivariate twin analyses were fit to simultaneously evaluate the influences of A, C, and E on child-report irritability and symptoms of depression, panic, generalized anxiety, separation anxiety, and social phobia. Like traditional factor analysis, this analysis accounts for covariation among multiple outcomes; however, twin data allows further decomposition of genetic and environmental contributions to the covariances among irritability and IDs. These analyses compared analytic models examining the structural relationship of those etiologic components distributed among the outcome syndromes. First, we applied a saturated model based on a Cholesky decomposition where all latent sources of A, C, and E variances and covariances are estimated. This also estimates the correlation of genetic and environmental factors contributing to each pair of outcomes (see Table 1).

Table 1.

Within-Person Phenotypic (r_P) and Twin Pair-Derived Genetic (r_A), Shared Environmental (r_C), and Unique Environmental (r_E) Correlations between Irritability, Depression, and Each Anxiety Symptom Domain (N = 355 twin pairs)

	Irritability	Depression	GAD	Separation	Social	Panic
Mean (SD)	2.85 (2.68)	5.23 (4.17)	5.94 (3.70)	5.20 (3.43)	6.00 (3.29)	5.17 (3.86)
rP	Irritability	Depression	GAD	Separation	Social	Panic
Irritability	1
Depression	0.42	1
GAD	0.32	0.51	1
Separation	0.33	0.39	0.47	1
Social	0.27	0.32	0.40	0.47	1
Panic	0.36	0.49	0.52	0.50	0.41	1
rA	Irritability	Depression	GAD	Separation	Social	Panic
Irritability	1
Depression	0.27	1
GAD	0.30	0.64	1
Separation	0.33	0.53	0.41	1
Social	−0.19	−0.22	0.03	0.15	1
Panic	0.06	0.32	0.62	0.60	−0.29	1
rC	Irritability	Depression	GAD	Separation	Social	Panic
Irritability	1
Depression	1	1
GAD	0.65	0.68	1
Separation	0.65	0.68	1	1
Social	0.94	0.95	0.88	0.88	1
Panic	0.96	0.97	0.83	0.83	1	1
rE	Irritability	Depression	GAD	Separation	Social	Panic
Irritability	1
Depression	0.36	1
GAD	0.28	0.40	1
Separation	0.25	0.24	0.38	1
Social	0.23	0.32	0.45	0.42	1
Panic	0.30	0.42	0.42	0.35	0.39	1

Note. r_p indicates phenotypic correlation; r_A indicates correlation of genetic components (A); r_C indicates correlation of shared environmental components (C); r_E indicates correlation of nonshared environmental components (E); GAD = symptoms of generalized anxiety disorder; Separation indicates symptoms of separation anxiety; Social indicates symptoms of social phobia; panic indicates symptoms of panic disorder. A histogram of irritability severity is provided in Supplemental Figures 1 and 2.

To clarify the structure of genetic and environmental influences, we fit the common pathway (CPM) and independent pathway (IPM) models, which were compared to the Cholesky decomposition. In the CPM, common A, C, and E influences are assumed to generate variation in, and covariation between, outcomes through a common latent factor, which loads each outcome. Additionally, residual As, Cs, and Es factors account for syndrome-specific influence independent of the common factor. In the IPM, variation in, and covariation between, syndromes is decomposed to common latent Ac, Cc and Ec factors, which directly influence all outcomes, and syndrome-specific As, Cs, and Es factors. Some prior research suggests that more than one shared genetic factor may influence comorbidity of IDs in adults (Hettema et al., 2006); therefore we compared the fit of two IPMs containing one or two Ac factors, respectively.

Selection of an optimal model was based on chi-square difference tests and comparatively low Akaike Information Criteria (AIC). The significance of common and specific A, C, and E components was evaluated using the chi-square difference test of worsened model fit when each component is constrained to zero (Neale & Cardon, 1992). Analyses were conducted using the OpenMx package (Neale et al., 2016) in R (R Core Team, 2015). Confidence intervals were estimated using profile-based maximum likelihood (e.g., Pritikin, Rappaport, & Neale, 2017). Additionally, given the established association of irritability and IDs with sex and age, all models included their effects on the means of irritability and each internalizing syndrome.

Results

Irritability

In the univariate analysis of irritability using child-report ARI, additive genetic (A) influences accounted for 34.4% (95% CI = 0–47.0%), and nonshared environmental influences (E) accounted for 65.6% (95% CI = 53.0–82.3%), of variance in irritability, respectively, whereas shared familial environmental (C) factors did not account for variance in irritability. Testing nested models to evaluate significance of each parameter suggests that the C component could be removed without change in model fit (p = 1.00). The resulting AE sub-model provided a significant additive genetic influence (heritability) of 34.4% (95% CI = 20.3–47.0%). We obtained a similar estimate of heritability (i.e., A) using the child- and parent-report ARI as indicators of a latent irritability phenotype of 35.1% (95% CI = 10.1–60.5%).

Biometric Analyses of Irritability with Depression and Anxiety Symptoms

Phenotypic Correlations.

The within-person correlations among irritability and each internalizing symptom domain were estimated within a multivariate biometrical twin model (top of Table 1). Irritability was significantly and moderately correlated with symptoms of depression (r = 0.42, 95% CI: 0.36, 0.48), panic (r = 0.36, 95% CI: 0.29, 0.42), generalized anxiety (r = 0.32, 95% CI: 0.25, 0.39), separation anxiety (r = 0.33, 95% CI: 0.26, 0.39), and social phobia (r = 0.27, 95% CI: 0.20, 0.34). Similarly, measures of anxiety and depression were also inter-correlated (r’s = 0.32–0.52).

Multivariate Twin Analyses.

The bottom three sections of Table 1 depict estimates for the bivariate genetic (rA), common environmental (rC), and unique environmental (rE) correlations, respectively. Several patterns are evident. First, descriptively, the genetic correlations of irritability with depression (r_A = 0.27), generalized anxiety (r_A = 0.30), and separation anxiety (r_A = 0.33) are moderate while irritability is modestly correlated with social phobia (r_A = −0.19) and panic (r_A = 0.06). Second, relative to genetic correlations with irritability, there are larger genetic correlations among depression, generalized anxiety, separation anxiety, and panic (see Table 1). Third, the shared environmental correlations are large and positive between all the measures (r_C’s = 0.65–1). Finally, unique environmental correlations between the measures are low to moderate (r_E’s = 0.23–0.45).

Table 2 compares the fit of the competing multivariate models. By the criteria of minimal AIC and chi-square difference test, the IPM with one common factor for each domain A, C, and E fits the data best. Detailed significance testing (Table 3) of each model component suggests that syndrome-specific C influences (Cs) can be removed in favor of retaining the common C factor (Cc).

Table 2.

Model Fit Indices of Competing Multivariate Twin Models of Irritability and Depression and Anxiety Symptoms (N = 355 twin pairs)

Model	np	−2LL	df	AIC	ΔLL	Δdf	p
Cholesky	81	21052.11	4149	12754.10	--	--	--
IPM 1A 1C 1E	53	21074.99	4177	12720.99	22.89	28	0.74
IPM 2A 1C 1E	59	21071.05	4171	12729.05	18.94	22	0.65
CPM	45	21111.10	4186	12739.10	59.00	37	0.01

Note. IPM = Independent Pathway Model; CPM = Common Pathway Model; A=Additive Genetic, C=Shared Environment, E=Non-shared Environment, np = number of parameters, LL=log-likelihood, AIC=Akaike’s Information Criterion, df=degrees of freedom. Bolded row indicates the best fitting model according to AIC.

Table 3.

Significance Testing for Components of 1A 1C 1E Independent Pathway Model

Model Description	Variance Components	np	−2LL	df	AIC	ΔLL	Δdf	p
Original Model…	1A 1C 1E - as, cs, es	53	21074.99	4177	12720.99	--	--	--
…Without Regression of Mean on Age	1A 1C 1E - as, cs, es	47	21175.09	4183	12809.09	100.10	6	< 0.0001
…Without Regression of Mean on Sex	1A 1C 1E - as, cs, es	47	21091.95	4183	12725.95	16.96	6	0.009
…Without All A Components	1C 1E - cs, es	42	21129.44	4188	12753.44	54.45	11	< 0.0001
…Without All C Components	1A 1E - as, es	41	21105.88	4189	12727.88	30.89	12	0.002
…Without All A & C Components	1E - es	30	21262.89	4200	12862.89	187.90	23	< 0.0001
…Without Common A Pathway	1C 1E - as, cs, es	48	21106.57	4182	12742.57	31.58	4	< 0.0001
…Without Common C Pathway	1A 1E - as, cs, es	47	21105.88	4183	12739.88	30.89	6	< 0.0001
…Without Common E Pathway	1A 1C - as, cs, es	47	21258.94	4183	12892.94	183.95	6	< 0.0001
…Without Specific A Components	1A 1C 1E – cs, es	47	21095.04	4183	12729.04	20.05	6	0.003
…Without Specific C Components	1A 1C 1E – as, es	47	21074.99	4183	12708.99	0	6	1.00

Note. A/a=Additive Genetic Components, C/c=Shared Environment Components, E/e=Non-shared Environment Components; np = number of parameters, LL=log-likelihood, AIC=Akaike’s Information Criterion, df=degrees of freedom. Bolded row indicates the best fitting model according to AIC.

Overall variance components for each outcome and estimated loadings for the final IPM are listed in Table 4 and depicted in Figure 1. We note several findings. First, E represents the overall largest proportion of variance (50–67 %) for all measures. Second, genetic and environmental influences operate, at least partially, through a pathway common to irritability and each ID studied. This “generalist” pathway accounts for the significant contribution of shared environmental contributions to irritability and IDs (Table 3). Finally, irritability and each ID are also influenced by syndrome-specific genetic and unique environmental contributions.

Table 4.

Total Variance Components and Path Loadings from Best-Fit Independent Pathway Model

	Overall Variance Components			Common (c) Loadings		Specific (s) Loadings
	A	C	E	ac	cc	ec	as	es
Irritability	0.32 (0.20–0.43)	0.05 (0–0.12)	0.64 (0.52–0.76)	0.37	0.21	0.32	0.42	0.73
Depression	0.43 (0.28–0.56)	0.07 (0–0.18)	0.50 (0.40–0.61)	0.64	0.26	0.39	0.16	0.59
Generalized Anxiety	0.34 0.21–0.47)	0.12 (0.04–0.21)	0.55 (0.44–0.67)	0.35	0.34	0.52	0.47	0.53
Panic	0.22 (0.10–0.36)	0.17 (0.07–0.28)	0.61 (0.50–0.72)	0.31	0.41	0.50	0.36	0.60
Social Phobia	0.23 (0.10–0.34)	0.16 (0.07–0.26)	0.62 (0.51–0.74)	0.05	0.39	0.49	0.47	0.62
Separation Anxiety	0.10 (0–0.25)	0.29 (0.17–0.42)	0.61 (0.51–0.72)	0.14	0.54	0.47	0.28	0.62

A/a=Additive Genetic Variance/Pathway Loadings, C/c=Shared Environment Variance/Pathway Loadings, E/e=Non-shared Environment Variance/Pathway Loadings

Figure 1.

Illustration of Best-Fitting Etiologic Model for Childhood Irritability and Internalizing Syndromes, Note. IRR = irritability; DEP = depression symptoms; GAD = generalized anxiety symptoms; PAN = panic symptoms; SOC = social phobia symptoms; SEP = separation anxiety symptoms. Ac, Cc, and Ec represent common genetic (A), shared environment (C), and unique (i.e., non-shared) environmental latent factors (E), respectively, while As and Es represent syndrome-specific genetic and unique environmental latent factors.

Supplemental Analyses: Biometric Analyses of Irritability with Externalizing Syndromes

Similar to analyses for internalizing syndromes (e.g., depression), within-person correlations among irritability and externalizing syndromes were estimated within a multivariate biometrical model (see Supplemental Table 1). Since externalizing syndromes were assessed only by parent-report, models of the association with externalizing syndromes used parent-report of irritability. Irritability is moderately correlated with attention deficit/hyperactivity disorder, oppositional defiant disorder, and conduct disorder, which is predominantly due to correlated genetic influences (see rA in Supplemental Table 1). Similar to internalizing syndromes, comparison of alternative multivariate models indicates that genetic and environmental contributions to irritability and externalizing syndromes in children are best summarized by the IPM (see Supplemental Table 2). Subsequent, nested models demonstrate that irritability and externalizing syndromes are robustly influenced by a common genetic and nonshared (i.e., person-specific) environmental pathway as well as disorder-specific genetic and environmental influences (see Supplemental Table 3). There is no evidence that shared (i.e., family-level) environment (i.e., C) contributes to the association of irritability with externalizing syndromes. However, as previously discussed, genetic influences may be overestimated when using parent-report in cases where parents report on both twins. Estimates from the best-fitting model illustrate that common genetic and nonshared environmental influences contribute to irritability and externalizing syndromes although there is additional evidence of disorder-specific genetic and nonshared environmental contributions (see Supplemental Table 4).

Follow-up analyses sought to examine the correlation between common genetic and nonshared environmental contributions to internalizing and externalizing syndromes. Two separate independent pathway models were fit to model common genetic and nonshared environmental influences to internalizing and externalizing syndromes. Parameters were fit to examine the correlation of common genetic factors (i.e., Ac) and of common nonshared environmental factors (i.e., Ec) between both independent pathway models. Based on prior evidence, both internalizing and externalizing models included irritability. Since externalizing syndromes were only assessed by parent-report, parent-report on all variables was used in these analyses. A chi-square difference test indicated that all shared (i.e., family-level) environmental influences (i.e., C) could be removed without worsening model fit, χ² (20) = 28.56, p = 0.10. There is no evidence of a correlation between the genetic influences to internalizing and externalizing syndromes, r = −0.04, p = 0.53. However, after removing this nonsignificant genetic correlation, irritability loaded significantly onto the genetic pathways for both internalizing, λ = 0.07, p = 0.04, and externalizing, λ = 0.18, p = 0.004, syndromes. Regarding nonshared environmental influences, irritability loads only onto the common environmental pathway for externalizing syndromes, λ = 0.73, p ≤ 0.0001; however, the common nonshared environmental pathway (i.e., Ec) for externalizing syndromes is highly correlated with that for internalizing syndromes, r = 0.57, p ≤ 0.0001.

Discussion

This study, conducted in 748 twin children aged 9 to 14, attempted to address four primary questions, the findings of which we review in turn.

What is the heritability of childhood irritability as assessed by the ARI? To our knowledge, this is the first twin study that assessed irritability using an instrument designed explicitly for that purpose. The univariate heritability estimate obtained for irritability using the child-report version of the ARI was 0.33 with the remaining variance accounted for by non-shared environment. Similar results, an estimated heritability of 0.35, arose from an analysis using both child- and parent-reports as indicators of a latent irritability factor. We note that when irritability was analyzed in the multivariate model, inclusion of internalizing syndromes allowed detection of a significant, shared environmental component (C) that slightly reduced the heritability estimate.

Our estimate is consistent with heritability reported in a twin/sibling study of older children and adolescents (i.e., 0.31; Stringaris, Zavos, et al., 2012) using an irritability score derived from child-report on oppositionality items of the Achenbach System of Empirically Based Assessment (ASEBA; Achenbach & Rescorla, 2003). While the present study demonstrated a consistent estimate of heritability from a latent composite of child- and parent-reported irritability, this is somewhat lower than was found at each of four child/adolescent ages in another study that derived an irritability score from solely parent-report on CBCL or ABCL oppositionality items (0.50–0.56; Savage et al., 2015). As discussed above, higher estimated heritability may be due to the use of parent-report in Savage et al. (2015) whereas the current study and Stringaris and colleagues (2012), who relied on participant self-report of irritability, found consistent yet lower estimated heritability.

Is irritability associated with all childhood internalizing syndromes? Irritability was moderately correlated with each childhood internalizing syndrome including those with primary affective (e.g., generalized anxiety) vs. fear-related (e.g., panic) components. This clarifies prior findings of moderate levels of irritability in youth with various anxiety disorders (Stoddard et al., 2014) and confirms the importance of irritability as a transdiagnostic element across childhood internalizing disorders.

What role do the genetic and environmental factors underlying childhood irritability play in risk for depression versus the various anxiety syndromes? As indicated in Tables 1 and 4, irritability showed moderately overlapping genetic risk with symptoms of generalized anxiety, separation anxiety, and depression (r_A’s = 0.27 – 0.33) but weaker correlation with symptoms of panic or social phobia (r_A’s = −0.19 – 0.06). Overall larger shared environmental correlations (r_C’s = 0.65 – 1) were seen between irritability and the internalizing symptoms, while unique environmental correlations tended to be smaller (r_E’s = 0.23 – 0.36). As illustrated in Figure 1, environmental factors underlying irritability accounted for the largest proportion of shared risk between it and the various anxiety domains with additional elements of moderate shared genetic risk primarily for depression, generalized anxiety and separation anxiety.

Do irritability-related etiological factors account for the observed comorbidity and overlap in risk between depression and anxiety syndromes? As indicated in Table 1, there was substantial genetic and environmental inter-correlation amongst anxiety syndromes and depression. Moreover, as evidenced by the IPM, a common set of genetic and environmental factors influence the etiology of irritability and each internalizing syndrome (Table 3 and Figure 1). Shared (i.e., familial) and unique (i.e., twin-specific) environmental factors influenced irritability and internalizing syndromes through a common pathway that influenced all syndromes. Additionally, evidence of specific genetic and unique environmental factors suggests potential syndrome-specific etiologic influences. Thus, genetic and environmental factors underlying childhood irritability play a significant but only partial role in explaining the development of comorbidity within internalizing psychopathology. Furthermore, the nearly complete overlap in genetic risk between major depression and generalized anxiety in adults (Hettema et al., 2006) is attenuated within the corresponding childhood syndromes. This could be due to neurodevelopmental changes that emerge later in adolescence, differences in assessment, or both.

We note that these results partially differ from those reported in a prior twin study of irritability and depression (Stringaris, Zavos, et al., 2012). Both the present and prior studies estimated comparable heritability of irritability and phenotypic correlations between irritability and depression (0.45 and 0.42, respectively). However, the prior study reported a higher genetic correlation between irritability and depression (r_A=0.70) than the present study (r_A=0.27). There are several potential explanations, which clarify the contribution of genetic and environmental factors to irritability and internalizing syndromes in childhood and adolescence.

First, Stringaris et al. studied an adolescent and young adult sample (14–23 years; mean=15 years). A longitudinal twin study has suggested age-related changes in the genetic correlation between irritability and mixed anxious/depressed symptoms (Savage et al., 2015). The present study suggests that the correlation of irritability with depression in childhood may be influenced more heavily by common familial (i.e., shared) environmental factors (r_C=0.67) than in adolescence or young adulthood. This is consistent with propositions that familial genetic and environmental influences may change with age (Bergen, Gardner, & Kendler, 2007). Second, the present study examined multiple ID phenotypes, which increases power to detect common etiological effects (e.g., C_C) compared to models of fewer phenotypes (Neale & Cardon, 1992). Additionally, final models in Stringaris et al. (2012) did not allow for shared environmental influences. We note that the estimated correlation for unique environmental effects is comparable between studies (r_E = 0.29, 0.21–0.37). Therefore, the present study provides new insights that familial environmental influences contribute to the correlation of irritability and depression which previously appeared to be entirely genetic in origin. Finally, while both studies used the same instrument to assess depression, the methods used to assess irritability differed. Prior research, including Stringaris et al., used an irritability composite created from items derived from scales designed to assess externalizing psychopathology (e.g., oppositionality). However, those items are conceptually distinct from the current conceptualization of irritability (Vidal-Ribas et al., 2016) and items designed to specifically assess irritability, such as the ARI.

In supplemental analyses, similar to internalizing psychopathology, we identify large correlations of irritability with externalizing psychopathology (i.e., ADHD, oppositional defiant disorder, and conduct disorder; see Supplemental Table 1), which is consistent with conceptual overlap of irritability and externalizing psychopathology. While the correlation of irritability with externalizing psychopathology is largely due to common genetic and person-specific environmental influences (see Supplemental Table 2), we demonstrate that some genetic and environmental influences are disorder specific (see Supplemental Tables 3 and 4). Finally, we demonstrate that, when accounting for both, irritability is associated with internalizing and externalizing psychopathology due to correlated environmental but different genetic pathways.

These findings should be considered in the context of several potential limitations. First, irritability, internalizing, and externalizing syndromes were assessed at the same time point. While the twin method provides insights into potentially causal etiological pathways, subsequent longitudinal research is needed to clarify the associations among dynamic genetic and environmental influences to irritability and internalizing syndromes (e.g., Roberson-Nay et al., 2015). For example, the present study was insufficiently powered to examine the influence of age and biological sex on genetic covariation of irritability with internalizing syndromes (Verhulst, 2017). However, the present study provides a foundation to support further work in a larger sample to assess the influence of age and sex on the heritability of irritability and its genetic covariation with internalizing syndromes (e.g., anxiety disorders). Second, the present study evaluated the relationship of irritability with internalizing and externalizing syndromes rather than disorders. Based on clinician assessment, the VCU-JAS sample is representative of child community samples, though low prevalence of IDs makes generalization of these findings to patient samples uncertain. However, research on the development of IDs requires understanding the etiology of dimensional risk factors in community samples (Wakschlag et al., 2015), and the SCARED, SMFQ, and Child Behavior Checklist index clinical risk.

Third, while the present study provides a representative community sample of 748 child twins, its size is smaller than some previously used to study genetics of irritability, thus limiting our statistical power. However, those prior studies did not examine the broad spectrum of specific internalizing syndromes as did the present one. A larger sample size would permit more precise estimates of genetic and environmental contributions to etiology of irritability and specific IDs. Statistically significant evidence of common and syndrome-specific genetic and environmental contributions to irritability and IDs, given limited statistical power, further indicates the robustness of these effects. Fourth, results regarding the association of irritability with externalizing syndromes warrant considerable caution due to possible inflation of genetic estimates due to parent-report about both twins. As detailed above, zygosity indicates difficulty distinguishing between children. Therefore, along with well-documented rater effects (Kenny et al., 2006), parents who indicated that their children are monozygotic twins may rate their children as more similar to each other than parents of dizygotic twins. The inclusion of additional informants would strengthen the assessment of irritability and psychopathology; however, future research should include informants familiar with only one twin to avoid confounds based on twin resemblance and rater effects. Finally, to facilitate future molecular genetic research, only Caucasian families were included. Further research is needed to generalize the present findings to other racial and ethnic groups.

Evidence of shared genetic and environmental factors underlying irritability and internalizing symptomatology in children has important implications for future research into risk pathways to anxiety and depressive disorders. Irritability has been proposed as a developmental risk marker for anxiety and depressive disorders that can be assessed earlier and more simply than clinical symptomatology (Vidal-Ribas et al., 2016). The current results confirm and extend the etiological basis for such a strategy. Additionally, the present study demonstrates predominantly similar common and disorder-specific genetic and environmental contributions to irritability and externalizing syndromes. While genetic and environmental contributions to irritability are associated with internalizing and externalizing syndromes, when accounting for the mutual contribution of irritability there is no evidence of a correlation of the genetic contribution to internalizing syndromes with that to externalizing syndromes.

The search for novel susceptibility genetic variants for IDs may benefit from considering childhood irritability as a useful genetic endophenotype (Savage, Sawyers, Roberson-Nay, & Hettema, 2017). The temporally changing landscape of childhood internalizing syndromes and relatively low prevalence of diagnostic-level disorders makes it difficult to profitably use childhood clinical samples for genetic discovery efforts. However, irritability is a behavioral dimension that can be reliably assessed in large epidemiological samples. Except as a component of bipolar disorder in adults (Greenwood, Akiskal, Akiskal, & Kelsoe, 2012) or attention-deficit hyperactivity disorder in youths (Aebi et al., 2016), no genome-wide association studies of irritability have been conducted. The identification of genetic variants underlying individual differences in childhood irritability could provide novel insights into the notably complex and heterogeneous genetic architectures of anxiety and depressive disorders that have thus far complicated gene-finding efforts (e.g., Shimada-Sugimoto, Otowa, & Hettema, 2015).

Finally, our results suggest that childhood irritability may be more informatively included as a specific predictor in developmental psychopathology, further refining risk models already in use (e.g., Rice et al., 2017). Clarifying genetic subtypes and mechanistic pathways could eventually inform selection of clinical interventions and prognostic evaluation.

Conclusion

In conclusion, the present study demonstrates that, when assessed in the current conceptualization, irritability is heritable in children at a similar rate to prior studies in children and adolescents. Moreover, the present study replicates evidence that irritability is correlated with anxiety and depressive syndromes. Finally, the present study describes common genetic and environmental influences, including common familial environmental influences, that may explain the evident association of irritability with anxiety and depressive syndromes in childhood.

Supplementary Material

Supplemental Figure 1.

Histogram of Raw Irritability Score on the Affective Reactivity Index

Supplemental Figure 2.

Histogram of Residual Irritability Severity Note. Irritability severity is based on self-report on the Affective Reactivity Index regressed on participant age and sex.

Supplemental Table.

Histogram of Residual Irritability Severity Note. Irritability severity is based on self-report on the Affective Reactivity Index regressed on participant age and sex.