0
Get Alert
Please Wait... Processing your request... Please Wait.
You must sign in to sign-up for alerts.

Please confirm that your email address is correct, so you can successfully receive this alert.

1
Article   |    
Association of Autistic-Like and Internalizing Traits During Childhood: A Longitudinal Twin Study
Victoria Hallett, M.Sc.; Angelica Ronald, Ph.D.; Fruhling Rijsdijk, Ph.D.; Francesca Happé, Ph.D.
Am J Psychiatry 2010;167:809-817. doi:10.1176/appi.ajp.2009.09070990
View Author and Article Information

The authors report no financial relationships with commercial interests.

The Twins' Early Development Study is funded by a program grant from the United Kingdom Medical Research Council (grant no. G500079).

Previously presented as a poster at the International Meeting for Autism Research, Chicago, May 7—9, 2009. Received July 15, 2009; revisions received Oct. 13 and Dec. 3, 2009; accepted Dec. 7, 2009. From the Social, Genetic, and Developmental Psychiatry Centre, Institute of Psychiatry, Kings College London, London; and the Centre for Brain and Cognitive Development, School of Psychology, Birkbeck College, London. Address correspondence and reprint requests to Miss Hallett, Box P080, Social, Genetic, and Developmental Psychiatry Centre, Institute of Psychiatry, De Crespigny Park, London, SE5 8AF, UK; victoria.hallett@iop.kcl.ac.uk (e-mail).

Received July 15, 2009; Accepted December 7, 2009.

Copyright © American Psychiatric Association

Objective  Children with autism spectrum disorders often experience severe anxiety and depression, yet the explanation for this association remains unclear. The authors examined the longitudinal relationship between autistic-like and internalizing traits across middle to late childhood in a population-representative twin sample.

Method  Participants were approximately 6,000 twin pairs born in England and Wales from 1994 to 1996. Parental reports of autistic-like and internalizing traits were analyzed at ages 7 and 8 (timepoint 1) and again at age 12 (timepoint 2). The direction and etiology of the associations between these traits were examined within a cross-lagged design.

Results  Findings revealed an asymmetric bidirectional association between autistic-like and internalizing traits over time. Autistic-like traits at age 7 made a modest but significant contribution to the presence of internalizing traits at age 12. Earlier internalizing traits also influenced the development of later autistic-like traits, although this association was approximately one-half the magnitude. While both traits were moderately to highly heritable, they were largely independent with regard to their genetic influences. Stronger associations were found between the modest shared environmental influences on each trait. Of note, it was autistic-like communication difficulties, rather than social deficits, that made a significant contribution to later internalizing traits.

Conclusions  The association between autistic-like and internalizing traits was attributable to reciprocal processes occurring across childhood, suggesting that these traits may serve to exacerbate each other over time. Autistic-like communication difficulties had the most notable impact. This association must now be explored within samples of children with diagnosed autism spectrum disorders and internalizing disorders, since this may help to inform the best timing and targeting of clinical intervention.

Abstract Teaser
Figures in this Article

Autism spectrum disorders are diagnosed on the basis of social and communication difficulties and the presence of repetitive behaviors and interests. However, in addition to these core impairments, other emotional problems are often observed, including internalizing difficulties such as anxiety (1, 2) and depression (3). Recent meta-analyses have reported that up to 84% of children with autism spectrum disorders experience impairing anxiety (4) and up to 34% experience depression (3). However, despite this well-reported overlap, few studies have addressed the longitudinal links between these difficulties.

In the present study, the association between autistic-like and internalizing traits was explored across middle to late childhood. By examining traits within a population-representative sample, not solely the clinical extreme, we gained greater power for quantitative genetic analysis (5). First, we tested whether autistic-like traits at age 7 predicted the level of internalizing traits at age 12. Second, we explored the reverse association, whereby earlier internalizing traits influence the subsequent development of autistic-like traits. This direction of effect is in line with findings of increased autistic-like traits in individuals with internalizing disorders (6, 7), suggesting that internalizing traits might exacerbate autistic-like difficulties. Finally, we considered whether there are bidirectional processes underlying this association.

A genetically informative cross-lagged design was used to estimate the phenotypic-driven processes by which internalizing and autistic-like traits affect one another over time. This model constrains all associations to take the form of partial regression coefficients, estimating the strength of the longitudinal relationship between the traits while controlling for their preexisting association at the earlier timepoint. This is the first study, to our knowledge, to use the cross-lagged model to assess autistic-like traits, although it has been used previously to disentangle other longitudinal relationships in the behavioral problems literature (810). We also explored, phenotypically, the impact of particular types of autistic-like traits on later internalizing traits. That is, are there distinguishable effects of communication impairments, social difficulties, and repetitive behaviors on later internalizing difficulties?

Previous investigations using the Twins' Early Development Study (11) sample have revealed high heritabilities for autistic-like traits (12, 13), moderate heritabilities for internalizing traits (14), and only modest levels of genetic overlap between these two traits at ages 8 and 9 (15). Our study extended these findings by taking a longitudinal approach. We predicted that there would be a bidirectional relationship, with autistic-like traits earlier in childhood directly influencing the level of later internalizing traits and vice versa. We predicted a significant level of stability in both traits across childhood, with only modest levels of etiological overlap between them. We also predicted that all three types of autistic-like traits would contribute significantly to the variance of later internalizing difficulties.

+

Participants

Data were obtained from the Twins' Early Development Study (11), a population-representative sample of twins born in England and Wales from 1994 to 1996. Written informed consent was obtained from all parents. A total of 7,834 families returned the questionnaires when their twins were age 7, 6,762 families returned them when their twins were age 8, and 5,876 families returned questionnaires when their twins were age 12. Families were excluded if severe pre- or postnatal complications were reported or if one twin had a severe medical condition (see reference 16). This led to a total of 523 families being excluded from the age 7 twin cohort, 336 families from the age 8 twin cohort, and 200 families from the age 12 twin cohort. Table 1 summarizes the zygosity of the sample following these exclusions. Zygosity was determined using parent ratings of similarity (17), supplemented by DNA genotyping. The sample consisted of children across the full spectrum of autistic-like traits, including a small percentage of children (age 8: 1.2%; age 12: 1.8%) who scored above the cutoff point of 15 for a diagnosis of potential autism spectrum disorder according to the Childhood Autism Spectrum Test criteria (formerly known as the Childhood Asperger Syndrome Test [18, 19]). We refer to ages 7 and 8 as timepoint 1 and age 12 as timepoint 2.

 
Anchor for Jump
Table 1. Autistic-Like and Internalizing Traits Among Twin Boys and Girls in Middle to Late Childhood by Zygosity
+

Measures

Autistic-like traits. Autistic-like traits were assessed at ages 8 and 12 using the parent-reported Childhood Autism Spectrum Test (18). The test includes 31 key items (and six control items) scored with a "yes/no" response and is reported to be sensitive and specific (19), with good construct validity and test-retest reliability (20). The test showed strong internal consistency at ages 8 (alpha of 0.73) and 12 (alpha of 0.78). We also used specific subscales at age 8 to assess social difficulties (12 items: alpha of 0.56), communication difficulties (12 items: alpha of 0.64), and repetitive behaviors (seven items: alpha of 0.49).

Internalizing traits. Internalizing traits were measured at ages 7 and 12 using the emotional problems subscale of the Strengths and Difficulties Questionnaire (21). This subscale consists of five items answered using a three-point Likert scale (never=0; sometimes=1; often=2). Two items assess anxiety, two assess depression, and one measures somatic symptoms. The responses to the items showed moderate internal consistency at age 7 (alpha of 0.63) and age 12 (alpha of 0.68). Previous studies have supported the reliability and sensitivity of this questionnaire (22, 23) and have validated these items relative to those of the Child Behavior Checklist (24, 25) and Rutter questionnaires (26).

+

Statistical Analyses

+

Twin method

The twin method allowed us to determine the relative influences of additive genetic factors ("A"), shared environmental factors ("C"), and nonshared environmental factors ("E") on a trait (the ACE model). Shared environmental factors serve to make children within a family similar to each other, whereas nonshared environmental factors result in differences between siblings. The twin design assumes that monozygotic twin pairs share all of their genetic information, dizygotic twins share 50% of their genes on average, and all twins experience the same shared environmental influences (see reference 5).

Autistic-like and internalizing trait scores were calculated if >90% of the items were available. Prior to modeling, the scores were log transformed to reduce positive skew and regressed for age and sex according to standard procedures.

Statistical analyses were carried out using the structural equation modeling Mx program (27). The following three types of correlations were calculated: phenotypic correlations (rph: trait 1 in twin 1 correlated with trait 2 in twin 1), intraclass twin correlations (trait 1 in twin 1 correlated with trait 1 in twin 2), and cross-trait cross-twin correlations (trait 1 in twin 1 correlated with trait 2 in twin 2). These correlations were established at each timepoint as well as longitudinally.

+

Cross-Lagged Model

The cross-lagged model is shown in Figure 1. Cross-lagged associations functioned as partial regression coefficients to illustrate stability in autistic-like traits (b2,2) and internalizing traits (b1,1) and the contribution of autistic-like traits at timepoint 1 (ages 7 and 8) to internalizing traits at timepoint 2 (age 12) (b1,2) and vice versa (b2,1). The estimates take into account the preexisting association between the two traits at timepoint 1.

 
Anchor for JumpAnchor for Jump
Figure 1. The Longitudinal Cross-Lagged Model Used to Determine the Associations Between Autistic-Like and Internalizing Traits Across Middle to Late Childhooda

aIn this cross-lagged model, rectangles indicate the measured variables for internalizing and autistic-like traits. Circles indicate the following latent factors: A (additive genetic factors), C (shared environmental factors), and E (nonshared environmental factors). Timepoint 1=ages 7 and 8 years; timepoint 2=age 12 years. Standardized path estimates for these influences are represented (a, c, and e for each variable), along with genetic and environmental correlations (rA, rC, rE), longitudinal stability paths (b1,1, b2,2), and longitudinal cross-lagged paths (b2,1, b1,2).

The cross-lagged model estimated the relative contribution of genetic and environmental influences to individual traits at timepoints 1 and 2. For example, internalizing traits at timepoint 1 were influenced by genetic (a12), shared environmental (c12), and nonshared environmental (e12) factors. These estimates were also calculated at timepoint 2 (a32, c32, e32), taking into account the influences at timepoint 1.

The model also estimated the etiological influences on the covariation between the traits at both timepoints. For example, at timepoint 1 this included additive genetic (rA1), shared environmental (rC1), and nonshared environmental (rE1) correlations. A correlation of one suggested that the same influence affected both traits.

Transmission of etiological influences over time. The genetic and environmental influences on the variances of autistic-like and internalizing traits at age 12 could be broken down into influences that were specific to timepoint 2 and influences that were shared with autistic-like and internalizing traits at timepoint 1. For example, the total genetic variance of autistic-like traits at timepoint 2 could be analyzed as genetic influences that were shared with autistic-like traits at timepoint 1 (stability effects: b2,22×a22); specific to internalizing traits at timepoint 1 (cross-lagged effects: b1,22×a12); associated with the covariation between the two traits at timepoint 1 (common effects: 2×[b2,2×a2×rA1×a1×b1,2]); and specific to autistic-like traits at timepoint 2 (residual effects: a42).

Initially, all parameter estimates were allowed to differ for boys and girls, with only the values for genetic, shared environmental, and nonshared environmental correlations at both time points constrained to be equal across the sexes, according to standard procedure. Subsequently, parameter estimates and partial regression coefficients were equated for male and female subjects at timepoints 1 and 2. Each cross-lagged model was compared with a saturated model using likelihood ratio chi-square tests, with lower values of -2-log likelihood and Akaike information criterion reflecting a better fit.

Phenotypic Analyses of the Influence of Individual Autistic-Like Traits on Later Internalizing Traits

Pearson's correlations were carried out for the autistic-like traits subscale ratings at timepoint 1 and internalizing traits at timepoint 2. One twin per zygotic pair was selected at random to maintain independence among the cases.

Regression analyses were used to determine how much of the variance in internalizing traits at timepoint 2 could be attributed to each of the subscale ratings at timepoint 1, controlling for internalizing traits at timepoint 1. Univariate regressions were used to determine whether each individual subscale rating was significantly associated with later internalizing. If significant, each subscale rating was then included in a multiple regression.

Summary statistics are shown in Table 1. Scores at timepoint 2 were significantly lower than scores at timepoint 1 for both internalizing traits (t=11.76, df=5007, p<0.01) and autistic-like traits (t=6.90, df=4680, p<0.01). Boys scored significantly higher than girls on the Childhood Autism Spectrum Test at both timepoint 1 and timepoint 2. Conversely, girls scored significantly higher than boys for internalizing traits at timepoint 1 and timepoint 2. Zygosity had a significant effect on the mean scores for autistic-like traits, with dizygotic pairs scoring significantly higher than monozygotic twins at timepoint 1 and timepoint 2. A significant sex-by-zygosity interaction was observed for autistic-like traits at timepoint 2, with boys in opposite-sex pairs scoring higher than other male participants and girls in opposite-sex pairs scoring lower than other female participants.

+

Correlations

+

Phenotypic correlations

There was significant (p<0.01) stability over time in autistic-like traits (boys: rph=0.59; girls: rph=0.55) and internalizing traits (boys: rph=0.42; girls: rph=0.42). There were modest associations between autistic-like traits at timepoint 1 and internalizing traits at timepoint 2 (boys: rph=0.22; girls: rph=0.20) and between internalizing traits at timepoint 1 and autistic-like traits at timepoint 2 (boys: rph=0.24; girls: rph=0.20).

+

Intraclass twin correlations

As seen in Table 2, correlations were significantly higher for monozygotic twin pairs relative to dizygotic twin pairs for both traits at timepoints 1 and 2, indicating significant genetic influences. The correlations suggested low levels of genetic dominance effects for male autistic-like traits (the number of dizygotic male twin correlations was less than one-half the number of monozygotic male correlations). For girls, correlations indicated shared environmental influences for both traits (the number of dizygotic female twin correlations was more than one-half the number of monozygotic female correlations).

 
Anchor for Jump
Table 2. Twin Correlations Between Autistic-Like and Internalizing Traits Within and Cross-Age Among Boys and Girls in Middle to Late Childhooda
+

Cross-trait cross-twin correlations

At timepoints 1 and 2, there was a modest association between the autistic-like trait score for twin 1 and the internalizing trait score for twin 2. Monozygotic twin correlations were modest and less than double the dizygotic correlations, indicative of low levels of genetic overlap.

+

Longitudinal correlations

Monozygotic within-trait across-time correlations were moderate and higher than dizygotic correlations, suggesting a significant genetic influence on the stability of both traits over time.

+

Cross-Lagged Model

Fit statistics for the cross-lagged model are shown in Table 3. The model provided a worse fit for the data than the saturated model. This is a common finding when using multivariate models (8), attributable to the large number of constraints that are imposed. In large samples, the additive effects of small differences are picked up as significant, decreasing the overall fit.

 
Anchor for Jump
Table 3. Fit Statistics for the Longitudinal Cross-Lagged Modela

Male and female parameter estimates could be equated for both traits at timepoint 1, for internalizing traits at timepoint 2, and for the cross-lagged paths of coefficients b1,1 and b1,2 without significantly worsening the fit. Results from the best fitting model are shown in Figure 2.

 
Anchor for JumpAnchor for Jump
Figure 2. Cross-Lagged Model Analysis of Associations Between Internalizing and Autistic-Like Traits at Ages 7 and 8 (timepoint 1) and Age 12 (timepoint 2) Among Twin Boys and Girlsa

aThe following components are shown: additive genetic factors (A); shared environmental factors (C); and nonshared environmental factors (E). The correlations are as follows: additive genetic (rA); shared environmental (rC); and nonshared environmental (rE). Standardized path estimates for the additive genetic, shared environmental, and nonshared environmental factors are presented from the best fitting cross-lagged model. Path estimates at timepoint 1 are specific to age 12. Paths in the center of the diagram indicate longitudinal stability paths and longitudinal cross-lagged paths, with the thickness for the arrow depicting the relative magnitude. The dotted line indicates nonsignificant paths. Estimates are shown separately for boys and girls if they could not be equated within the best fitting model.

+

Cross-Lagged Associations

When each coefficient was fixed at 0, a deterioration in fit was observed (b2,1: ʔχ2=134.71 for Δ2df, p<0.01; b1,2: Δχ2=69.76 for Δ2df, p<0.01), suggesting that all cross-lagged paths were significant.

+

A similar pattern was observed for both sexes

First, there was a significant influence (p<0.01) of autistic-like traits at timepoint 1 on internalizing traits at timepoint 2 (boys: b2,1=0.15; girls: b2,1=0.10). Second, there was a significant, although more modest, influence (p<0.01) of internalizing traits at timepoint 1 on autistic-like traits at timepoint 2 (b1,2=0.06). Third, there was a significant level of stability (p<0.01) over time for internalizing traits (b1,1=0.37) and particularly autistic-like traits (boys: b2,2=0.56; girls: b2,2=0.51). That is, the variance in autistic-like traits at age 12 was somewhat attributable to autistic-like traits at ages 7 and 8 (boys: b2,2=0.562 [31%]; girls: b2,2=0.512 [26%]) and a modest proportion of the variance of internalizing traits at timepoint 2 was explained by internalizing traits at timepoint 1 (14%).

+

Genetic and Environmental Influences

Univariate modeling showed that genetic dominance was nonsignificant for both traits. As such, the model estimated A, C, and E parameters only. Heritability estimates were high for autistic-like traits at timepoint 1 (73%), with moderate nonshared environmental effects (20%) and a small influence of shared environment (6%). All three parameters were significant, since confidence intervals did not overlap with 0. Figure 2 also presents the proportion of A, C, and E components specific to timepoint 2. As described previously, we could calculate the overall influences of A, C, and E components on each trait at timepoint 2 (e.g., the overall heritability of autistic-like traits at timepoint 2 is the sum of genetic influences shared with autistic-like traits at timepoint 1, genetic influences shared with internalizing traits at timepoint 1, genetic influences as a result of the association between the two traits at timepoint 1, and new effects at timepoint 2). The pattern of etiological influences at timepoint 2 was similar to timepoint 1 for the C component (boys: shared environmental estimate=0.07; girls: shared environmental estimate=0.21) and E component (boys: nonshared environmental estimate=0.29; girls: nonshared environmental estimate=0.27), although heritability estimates (component A) were slightly lower (boys: genetic estimate=0.65; girls: genetic estimate=0.52).

Internalizing traits were moderately heritable at timepoint 1 (0.45), with a moderate influence of nonshared environment (0.37) and a small but significant influence of shared environment (0.18). At timepoint 2, overall parameter estimates were similar for the A (0.49) and E (0.42) components, with a slightly lower estimate for the C component (0.08). There was a significant but modest genetic correlation between the traits (timepoint 1: rA=0.17; timepoint 2: rA=0.17) and low levels of nonshared environmental overlap (timepoint 1: rE=0.08; timepoint 2: rE=0.14). Shared environmental influences correlated more highly between the two traits at timepoint 1 (rC=1.00 [0.73 to 1.00], although there was a nonsignificant estimate of shared environmental influences at timepoint 2 (rC=0.22[—0.03 to 0.78]).

+

Transmission of Etiological Influences

A table detailing the transmission of etiological influences is available in the data supplement accompanying the online version of this article, since these results do not form a core focus of this study. Some genetic influences on internalizing traits at timepoint 2 were transmitted from timepoint 1 (12%—17%), while most were specific to timepoint 2 (78%—82%). Nonshared environmental influences on internalizing traits were also mainly specific to timepoint 2 (86%—88%), while some were transmitted from timepoint 1 (12%). A small proportion of shared environmental effects were transmitted from timepoint 1 (15%—25%), some were shared with autistic-like traits at timepoint 1 (0.8%—2%), and some were shared with the covariance between internalizing and autistic-like traits at timepoint 1 (8%—13%). For additional information, see the data supplement.

There was significant stability in the genetic influences of autistic-like traits over time, with 35%—37% of influences at timepoint 2 transmitted from timepoint 1. There were also new genetic influences specific to the later time point (61%—63%). A small proportion of nonshared environmental influences was transmitted over time (19%—21%), while others emerged at timepoint 2 (71%—79%). A modest proportion of shared environmental influences at timepoint 2 was explained by autistic-like traits at timepoint 1 (24%—29%), internalizing traits at timepoint 1 (0.5%—1.4%), and their covariance at timepoint 1 (5%—14%).

+

Phenotypic Analyses of the Influence of Individual Autistic-Like Traits on Later Internalizing Traits

Pearson's correlations showed that all three autistic-like traits subscales at timepoint 1 correlated significantly with internalizing traits at timepoint 2 (p<0.01). Communicative difficulties (r=0.27) and repetitive behaviors (r=0.15) correlated most strongly with later internalizing traits, whereas social difficulties showed a weaker association (r=0.08). There was a strong significant correlation between internalizing traits at timepoints 1 and 2 (r=0.46, p<0.01).

In univariate regressions, all three subscale ratings significantly predicted internalizing traits at timepoint 2, albeit weakly, and thus were included in the full model (social skills: standardized β=0.03, p=0.01; communication skills: β=0.15, p<0.01; repetitive behaviors: β=0.07, p<0.01). In the full model, internalizing traits at timepoint 1 accounted for 46% of the variance in later internalizing traits (p<0.01). Further to this, social skills ratings made no additional contribution to the variance of internalizing traits at timepoint 2, and repetitive behaviors accounted for only 3% of the variance. Communication difficulties at timepoint 1 showed the strongest association with internalizing traits at timepoint 2, accounting for 14% (p<0.01) of the variance, after controlling for earlier levels of internalizing traits.

To our knowledge, this represents the first study to address the longitudinal links between autistic-like traits and internalizing traits across childhood. Our results point to an asymmetric bidirectional relationship between these traits. Autistic-like traits at timepoint 1 (ages 7 and 8) had a direct, although modest, phenotypic influence on the development of internalizing traits at timepoint 2 (age 12). The reverse association was also significant, with internalizing traits having a significant, albeit smaller, impact on later autistic-like traits.

Our findings suggest that autistic-like traits in later childhood were driven in part by earlier internalizing difficulties. For example, anxieties in earlier childhood may reduce opportunities for developing effective social communication skills with peers. Furthermore, a child that is easily distressed may cling to routines and may use repetitive behaviors to soothe high levels of arousal (28). That said, the reverse association was approximately twice as strong, implying that early autistic-like difficulties may have a more pronounced impact on subsequent anxiety and depression than vice versa. For instance, children who struggle with social interactions and communication may shy away from group situations and may find it difficult to express themselves and seek support. Similarly, adhering to rigid routines and rituals may also become more stressful as children progress through school.

Internalizing and autistic-like traits showed significant stability across a 5-year period. This is in agreement with a study of 8- to 15-year-old children with and without autism spectrum disorders, which found significant stability in social difficulties over time (29). Similarly, studies have also shown stability in internalizing symptoms throughout childhood and adolescence (30, 31). In line with previous findings, we found a high heritability for autistic-like traits and moderate heritability for internalizing traits (12, 14, 32). However, although significant, there were only low levels of genetic overlap between the two traits. This finding is consistent with a previous study of the present sample (15), which found a similarly low level of genetic overlap when the twins were 9 years old. Therefore, despite reciprocal phenotypic processes underlying these traits, the overlap does not appear to be driven by substantial shared genetic influences.

In contrast, there was a stronger association between the shared environmental influences on the two traits, particularly at ages 7 and 8. This finding should be interpreted with some caution because of the large confidence intervals and modest influence of shared environmental factors on autistic-like traits. However, it suggests that environmental factors, such as parental and home influences, may be important. For example, the opportunities that parents provide for socialization may impact children's social and communication skills. Similarly, parents may model coping styles that influence how children react to stress. While it is difficult to speculate about the mechanisms underlying the overlap between these traits, a focus on these environmental factors, rather than a shared genetic pathway, may be beneficial, at least within the general population.

Although not central to our aims, the model also addressed the transmission of etiological influences on autistic-like and internalizing traits over time. There was moderate continuity in the genetic factors involved at both time points, most notably for autistic-like traits. However, the majority of genetic factors were age-specific, pointing to the emergence of novel sets of influences at different developmental stages. A similar pattern of specificity was observed for environmental influences, possibly reflecting the key changes in schools and friendships across this period. Recent studies of internalizing (31, 33) and autistic-like traits (29) have emphasized their dynamic nature, highlighting both stability and change across development.

Autistic-like traits affected internalizing traits, but which subtypes of autistic-like traits were driving this association? Our findings suggest that earlier communication difficulties contributed most strongly to later internalizing traits. It seems logical that communicative impairments may confuse a child's understanding of other people, limit self-expression, and reduce strategies for coping with worries and sadness. Conversely, social difficulties at timepoint 1 did not significantly predict later internalizing traits. This unexpected finding may point to a complex relationship between social motivation and ability (as measured by the Childhood Autism Spectrum Test) and anxiety and depression. For example, a child with elevated social autistic-like traits may also lack the insight to reflect on his or her difficulties and to worry about evaluation by peers. In addition, children with communication difficulties who have more social insight might suffer greater anxiety about their own difficulties.

Further research is required to explore how these population-based findings translate to clinical samples of children with autism spectrum disorders and internalizing disorders. Children with autism spectrum disorders frequently experience anxiety (4, 34), while children with internalizing disorders exhibit heightened autistic-like traits (6). As such, we can speculate that a similar reciprocal relationship may be observed in clinical groups. Our findings also highlight the importance of differentiating between the effects of different subtypes of autistic-like traits. This may help to both refine our understanding of internalizing difficulties in autism and to inform the most effective targets for intervention.

This study benefitted from the use of a large sample and incorporating the same measures at two ages. However, a number of limitations must be considered. First, the cross-lagged model incorporated only parental-report measures. To reduce rater bias and correlated measurement error, future studies should use additional informants and/or behavioral observation. Second, the internalizing traits measure revealed modest internal consistency and incorporated only five items. A more detailed measure could help capture a wider range of internalizing difficulties. Third, both of the traits analyzed are heterogeneous, both phenotypically and genetically (13, 14), and thus further investigation is needed to address the relationship between specific internalizing subtypes and specific autistic-like traits. Finally, there are also some limitations inherent to the twin design (5) and the cross-lagged model (8). The modest cross-lag associations suggest that there are additional factors (not included in the model) that influence these traits over time.

Our results provide support for a modest bidirectional relationship between autistic-like and internalizing traits across childhood, with a particularly prominent influence of autistic-like traits on subsequent internalizing. It is important to consider how these results might translate to different developmental stages and clinical groups. If autistic symptoms, particularly communication difficulties, serve to exacerbate internalizing symptoms, promoting early communication skills may help to reduce later internalizing difficulties. Similarly, we can speculate that attempts to reduce stress in childhood might be somewhat protective against the worsening of later social communication problems.

The authors thank Professor Robert Plomin and his team for the use of data from the Twins' Early Development Study. The authors also thank all of the families who participated in the study.

de Bruin  EI;  Ferdinand  RF;  Meester  S;  de Nijs  PF;  Verheij  F:  High rates of psychiatric co-morbidity in PDD—NOS.  J Autism Dev Disord 2007; 37:877—886
[CrossRef] | [PubMed]
 
Simonoff  E;  Pickles  A;  Charman  T;  Chandler  S;  Loucas  T;  Baird  G:  Psychiatric disorders in children with autism spectrum disorders: prevalence, comorbidity, and associated factors in a population-derived sample.  J Am Acad Child Adolesc Psychiatry 2008; 47:921—929
[CrossRef] | [PubMed]
 
Stewart  ME;  Barnard  L;  Pearson  J;  Hasan  R;  O'Brien  G:  Presentation of depression in autism and Asperger Syndrome: a review.  Autism 2006; 10:103—116
[CrossRef] | [PubMed]
 
White  SW;  Oswald  D;  Ollendick  T;  Scahill  L:  Anxiety in children and adolescents with autism spectrum disorders.  Clin Psychol Rev 2009; 29:216—229
[CrossRef] | [PubMed]
 
Plomin  R;  Defries  J;  McClearn  GE;  McGuffin  P:  Behavioral Genetics, 5th ed. Edited by Plomin  R.  New York,  Worth Publishers, 2008
 
Pine  DS;  Guyer  AE;  Goldwin  M;  Towbin  KA;  Leibenluft  E:  Autism spectrum disorder scale scores in pediatric mood and anxiety disorders.  J Am Acad Child Adolesc Psychiatry 2008; 47:652—661
[CrossRef] | [PubMed]
 
Ivarsson  T;  Melin  K:  Autism spectrum traits in children and adolescents with obsessive-compulsive disorder (OCD).  J Anxiety Disord 2008; 22:969—978
[CrossRef] | [PubMed]
 
Burt  SA;  McGue  M;  Krueger  RF;  Iacono  WG:  How are parent-child conflict and childhood externalizing symptoms related over time?: Results from a genetically informative cross-lagged study.  Dev Psychopathol 2005; 17:145—165
[CrossRef] | [PubMed]
 
Larsson  H;  Viding  E;  Rijsdijk  FV;  Plomin  R:  Relationships between parental negativity and childhood antisocial behavior over time: a bidirectional effects model in a longitudinal genetically informative design.  J Abnorm Child Psychol 2008; 36:633—645
[CrossRef] | [PubMed]
 
Gregory  AM;  Rijsdijk  FV;  Lau  JY;  Dahl  RE;  Eley  TC:  The direction of longitudinal associations between sleep problems and depression symptoms: a study of twins aged 8 and 10 years.  Sleep 2009; 32:189—199
[PubMed]
 
Oliver  BR;  Plomin  R:  Twins' Early Development Study (TEDS): a multivariate, longitudinal genetic investigation of language, cognition and behavior problems from childhood through adolescence.  Twin Res Hum Genet 2007; 10:96—105
[CrossRef] | [PubMed]
 
Ronald  A;  Happe  F;  Bolton  P;  Butcher  LM;  Price  TS;  Wheelwright  S;  Baron-Cohen  S;  Plomin  R:  Genetic heterogeneity between the three components of the autism spectrum: a twin study.  J Am Acad Child Adolesc Psychiatry 2006; 45:691—699
[CrossRef] | [PubMed]
 
Ronald  A;  Happe  F;  Price  TS;  Baron-Cohen  S;  Plomin  R:  Phenotypic and genetic overlap between autistic traits at the extremes of the general population.  J Am Acad Child Adolesc Psychiatry 2006; 45:1206—1214
[CrossRef] | [PubMed]
 
Hallett  V;  Ronald  A;  Rijsdijk  F;  Eley  TC:  Phenotypic and genetic differentiation of anxiety-related behaviors in middle childhood.  Depress Anxiety 2009; 26:316—324
[CrossRef] | [PubMed]
 
Hallett  V;  Ronald  A;  Happe  F:  Investigating the association between autistic-like and internalizing traits in a community-based twin sample.  J Am Acad Child Adolesc Psychiatry 2009; 48:618—627
[CrossRef] | [PubMed]
 
Ronald  A;  Simonoff  E;  Kuntsi  J;  Asherson  P;  Plomin  R:  Evidence for overlapping genetic influences on autistic and ADHD behaviours in a community twin sample.  J Child Psychol Psychiatry 2008; 49:535—542
[CrossRef] | [PubMed]
 
Price  TS;  Freeman  B;  Craig  I;  Petrill  SA;  Ebersole  L;  Plomin  R:  Infant zygosity can be assigned by parental report questionnaire data.  Twin Res 2000; 3:129—133
[CrossRef] | [PubMed]
 
Scott  FJ;  Baron-Cohen  S;  Bolton  P;  Brayne  C:  The CAST (Childhood Asperger Syndrome Test): preliminary development of a UK screen for mainstream primary-school-age children.  Autism 2002; 6:9—31
[CrossRef] | [PubMed]
 
Williams  J;  Scott  F;  Stott  C;  Allison  C;  Bolton  P;  Baron-Cohen  S;  Brayne  C:  The CAST (Childhood Asperger Syndrome Test): test accuracy.  Autism 2005; 9:45—68
[CrossRef] | [PubMed]
 
Williams  J;  Allison  C;  Scott  F;  Stott  C;  Bolton  P;  Baron-Cohen  S;  Brayne  C:  The Childhood Asperger Syndrome Test (CAST): test-retest reliability.  Autism 2006; 10:415—427
[CrossRef] | [PubMed]
 
Goodman  R:  Psychometric properties of the Strengths and Difficulties Questionnaire.  J Am Acad Child Adolesc Psychiatry 2001; 40:1337—1345
[CrossRef] | [PubMed]
 
Muris  P;  Meesters  C;  van den Berg  F:  The Strengths and Difficulties Questionnaire (SDQ): further evidence for its reliability and validity in a community sample of Dutch children and adolescents.  Eur Child Adolesc Psychiatry 2003; 12:1—8
[CrossRef] | [PubMed]
 
Rothenberger  A;  Woerner  W:  Strengths and Difficulties Questionnaire (SDQ): evaluations and applications.  Eur Child Adolesc Psychiatry 2004; 13(suppl 2):II1—II2
[PubMed]
 
Klasen  H;  Woerner  W;  Wolke  D;  Meyer  R;  Overmeyer  S;  Kaschnitz  W;  Rothenberger  A;  Goodman  R:  Comparing the German versions of the Strengths and Difficulties Questionnaire (SDQ—Deu) and the Child Behavior Checklist.  Eur Child Adolesc Psychiatry 2000; 9:271—276
[CrossRef] | [PubMed]
 
Goodman  R;  Scott  S:  Comparing the Strengths and Difficulties Questionnaire and the Child Behavior Checklist: Is small beautiful? J Abnorm Child Psychol 1999; 27:17—24
[CrossRef] | [PubMed]
 
Goodman  R:  The Strengths and Difficulties Questionnaire: a research note.  J Child Psychol Psychiatry 1997; 38:581—586
[CrossRef] | [PubMed]
 
Neale  MC:  Mx: Statistical Modeling, 4th ed.  Charlottesville, Va,  Department of Psychiatry, Medical College of Virginia, 1997
 
Turner  M:  Annotation: repetitive behaviour in autism: a review of psychological research.  J Child Psychol Psychiatry 1999; 40:839—849
[CrossRef] | [PubMed]
 
Constantino  JN;  Abbacchi  AM;  Lavesser  PD;  Reed  H;  Givens  L;  Chiang  L;  Gray  T;  Gross  M;  Zhang  Y;  Todd  RD:  Developmental course of autistic social impairment in males.  Dev Psychopathol 2009; 21:127—138
[CrossRef] | [PubMed]
 
Ferdinand  RF;  Dieleman  G;  Ormel  J;  Verhulst  FC:  Homotypic versus heterotypic continuity of anxiety symptoms in young adolescents: evidence for distinctions between DSM-IV subtypes.  J Abnorm Child Psychol 2007; 35:325—333
[CrossRef] | [PubMed]
 
Kendler  KS;  Gardner  CO;  Lichtenstein  P:  A developmental twin study of symptoms of anxiety and depression: evidence for genetic innovation and attenuation.  Psychol Med 2008; 38:1567—1575
[CrossRef] | [PubMed]
 
Constantino  J;  Todd  RD:  Autistic traits in the general population: a twin study.  Arch Gen Psychiatry 2003; 60:524—530
[CrossRef] | [PubMed]
 
Haberstick  BC;  Schmitz  S;  Young  SE;  Hewitt  JK:  Contributions of genes and environments to stability and change in externalizing and internalizing problems during elementary and middle school.  Behav Genet 2005; 35:381—396
[CrossRef] | [PubMed]
 
Weisbrot  DM;  Gadow  KD;  DeVincent  CJ;  Pomeroy  J:  The presentation of anxiety in children with pervasive developmental disorders.  J Child Adolesc Psychopharmacol 2005; 15:477—496
[CrossRef] | [PubMed]
 
References Container

Figure 1.  The Longitudinal Cross-Lagged Model Used to Determine the Associations Between Autistic-Like and Internalizing Traits Across Middle to Late Childhooda

aIn this cross-lagged model, rectangles indicate the measured variables for internalizing and autistic-like traits. Circles indicate the following latent factors: A (additive genetic factors), C (shared environmental factors), and E (nonshared environmental factors). Timepoint 1=ages 7 and 8 years; timepoint 2=age 12 years. Standardized path estimates for these influences are represented (a, c, and e for each variable), along with genetic and environmental correlations (rA, rC, rE), longitudinal stability paths (b1,1, b2,2), and longitudinal cross-lagged paths (b2,1, b1,2).

Figure 2.  Cross-Lagged Model Analysis of Associations Between Internalizing and Autistic-Like Traits at Ages 7 and 8 (timepoint 1) and Age 12 (timepoint 2) Among Twin Boys and Girlsa

aThe following components are shown: additive genetic factors (A); shared environmental factors (C); and nonshared environmental factors (E). The correlations are as follows: additive genetic (rA); shared environmental (rC); and nonshared environmental (rE). Standardized path estimates for the additive genetic, shared environmental, and nonshared environmental factors are presented from the best fitting cross-lagged model. Path estimates at timepoint 1 are specific to age 12. Paths in the center of the diagram indicate longitudinal stability paths and longitudinal cross-lagged paths, with the thickness for the arrow depicting the relative magnitude. The dotted line indicates nonsignificant paths. Estimates are shown separately for boys and girls if they could not be equated within the best fitting model.

Anchor for Jump
Table 1. Autistic-Like and Internalizing Traits Among Twin Boys and Girls in Middle to Late Childhood by Zygosity
Anchor for Jump
Table 2. Twin Correlations Between Autistic-Like and Internalizing Traits Within and Cross-Age Among Boys and Girls in Middle to Late Childhooda
Anchor for Jump
Table 3. Fit Statistics for the Longitudinal Cross-Lagged Modela
+

References

de Bruin  EI;  Ferdinand  RF;  Meester  S;  de Nijs  PF;  Verheij  F:  High rates of psychiatric co-morbidity in PDD—NOS.  J Autism Dev Disord 2007; 37:877—886
[CrossRef] | [PubMed]
 
Simonoff  E;  Pickles  A;  Charman  T;  Chandler  S;  Loucas  T;  Baird  G:  Psychiatric disorders in children with autism spectrum disorders: prevalence, comorbidity, and associated factors in a population-derived sample.  J Am Acad Child Adolesc Psychiatry 2008; 47:921—929
[CrossRef] | [PubMed]
 
Stewart  ME;  Barnard  L;  Pearson  J;  Hasan  R;  O'Brien  G:  Presentation of depression in autism and Asperger Syndrome: a review.  Autism 2006; 10:103—116
[CrossRef] | [PubMed]
 
White  SW;  Oswald  D;  Ollendick  T;  Scahill  L:  Anxiety in children and adolescents with autism spectrum disorders.  Clin Psychol Rev 2009; 29:216—229
[CrossRef] | [PubMed]
 
Plomin  R;  Defries  J;  McClearn  GE;  McGuffin  P:  Behavioral Genetics, 5th ed. Edited by Plomin  R.  New York,  Worth Publishers, 2008
 
Pine  DS;  Guyer  AE;  Goldwin  M;  Towbin  KA;  Leibenluft  E:  Autism spectrum disorder scale scores in pediatric mood and anxiety disorders.  J Am Acad Child Adolesc Psychiatry 2008; 47:652—661
[CrossRef] | [PubMed]
 
Ivarsson  T;  Melin  K:  Autism spectrum traits in children and adolescents with obsessive-compulsive disorder (OCD).  J Anxiety Disord 2008; 22:969—978
[CrossRef] | [PubMed]
 
Burt  SA;  McGue  M;  Krueger  RF;  Iacono  WG:  How are parent-child conflict and childhood externalizing symptoms related over time?: Results from a genetically informative cross-lagged study.  Dev Psychopathol 2005; 17:145—165
[CrossRef] | [PubMed]
 
Larsson  H;  Viding  E;  Rijsdijk  FV;  Plomin  R:  Relationships between parental negativity and childhood antisocial behavior over time: a bidirectional effects model in a longitudinal genetically informative design.  J Abnorm Child Psychol 2008; 36:633—645
[CrossRef] | [PubMed]
 
Gregory  AM;  Rijsdijk  FV;  Lau  JY;  Dahl  RE;  Eley  TC:  The direction of longitudinal associations between sleep problems and depression symptoms: a study of twins aged 8 and 10 years.  Sleep 2009; 32:189—199
[PubMed]
 
Oliver  BR;  Plomin  R:  Twins' Early Development Study (TEDS): a multivariate, longitudinal genetic investigation of language, cognition and behavior problems from childhood through adolescence.  Twin Res Hum Genet 2007; 10:96—105
[CrossRef] | [PubMed]
 
Ronald  A;  Happe  F;  Bolton  P;  Butcher  LM;  Price  TS;  Wheelwright  S;  Baron-Cohen  S;  Plomin  R:  Genetic heterogeneity between the three components of the autism spectrum: a twin study.  J Am Acad Child Adolesc Psychiatry 2006; 45:691—699
[CrossRef] | [PubMed]
 
Ronald  A;  Happe  F;  Price  TS;  Baron-Cohen  S;  Plomin  R:  Phenotypic and genetic overlap between autistic traits at the extremes of the general population.  J Am Acad Child Adolesc Psychiatry 2006; 45:1206—1214
[CrossRef] | [PubMed]
 
Hallett  V;  Ronald  A;  Rijsdijk  F;  Eley  TC:  Phenotypic and genetic differentiation of anxiety-related behaviors in middle childhood.  Depress Anxiety 2009; 26:316—324
[CrossRef] | [PubMed]
 
Hallett  V;  Ronald  A;  Happe  F:  Investigating the association between autistic-like and internalizing traits in a community-based twin sample.  J Am Acad Child Adolesc Psychiatry 2009; 48:618—627
[CrossRef] | [PubMed]
 
Ronald  A;  Simonoff  E;  Kuntsi  J;  Asherson  P;  Plomin  R:  Evidence for overlapping genetic influences on autistic and ADHD behaviours in a community twin sample.  J Child Psychol Psychiatry 2008; 49:535—542
[CrossRef] | [PubMed]
 
Price  TS;  Freeman  B;  Craig  I;  Petrill  SA;  Ebersole  L;  Plomin  R:  Infant zygosity can be assigned by parental report questionnaire data.  Twin Res 2000; 3:129—133
[CrossRef] | [PubMed]
 
Scott  FJ;  Baron-Cohen  S;  Bolton  P;  Brayne  C:  The CAST (Childhood Asperger Syndrome Test): preliminary development of a UK screen for mainstream primary-school-age children.  Autism 2002; 6:9—31
[CrossRef] | [PubMed]
 
Williams  J;  Scott  F;  Stott  C;  Allison  C;  Bolton  P;  Baron-Cohen  S;  Brayne  C:  The CAST (Childhood Asperger Syndrome Test): test accuracy.  Autism 2005; 9:45—68
[CrossRef] | [PubMed]
 
Williams  J;  Allison  C;  Scott  F;  Stott  C;  Bolton  P;  Baron-Cohen  S;  Brayne  C:  The Childhood Asperger Syndrome Test (CAST): test-retest reliability.  Autism 2006; 10:415—427
[CrossRef] | [PubMed]
 
Goodman  R:  Psychometric properties of the Strengths and Difficulties Questionnaire.  J Am Acad Child Adolesc Psychiatry 2001; 40:1337—1345
[CrossRef] | [PubMed]
 
Muris  P;  Meesters  C;  van den Berg  F:  The Strengths and Difficulties Questionnaire (SDQ): further evidence for its reliability and validity in a community sample of Dutch children and adolescents.  Eur Child Adolesc Psychiatry 2003; 12:1—8
[CrossRef] | [PubMed]
 
Rothenberger  A;  Woerner  W:  Strengths and Difficulties Questionnaire (SDQ): evaluations and applications.  Eur Child Adolesc Psychiatry 2004; 13(suppl 2):II1—II2
[PubMed]
 
Klasen  H;  Woerner  W;  Wolke  D;  Meyer  R;  Overmeyer  S;  Kaschnitz  W;  Rothenberger  A;  Goodman  R:  Comparing the German versions of the Strengths and Difficulties Questionnaire (SDQ—Deu) and the Child Behavior Checklist.  Eur Child Adolesc Psychiatry 2000; 9:271—276
[CrossRef] | [PubMed]
 
Goodman  R;  Scott  S:  Comparing the Strengths and Difficulties Questionnaire and the Child Behavior Checklist: Is small beautiful? J Abnorm Child Psychol 1999; 27:17—24
[CrossRef] | [PubMed]
 
Goodman  R:  The Strengths and Difficulties Questionnaire: a research note.  J Child Psychol Psychiatry 1997; 38:581—586
[CrossRef] | [PubMed]
 
Neale  MC:  Mx: Statistical Modeling, 4th ed.  Charlottesville, Va,  Department of Psychiatry, Medical College of Virginia, 1997
 
Turner  M:  Annotation: repetitive behaviour in autism: a review of psychological research.  J Child Psychol Psychiatry 1999; 40:839—849
[CrossRef] | [PubMed]
 
Constantino  JN;  Abbacchi  AM;  Lavesser  PD;  Reed  H;  Givens  L;  Chiang  L;  Gray  T;  Gross  M;  Zhang  Y;  Todd  RD:  Developmental course of autistic social impairment in males.  Dev Psychopathol 2009; 21:127—138
[CrossRef] | [PubMed]
 
Ferdinand  RF;  Dieleman  G;  Ormel  J;  Verhulst  FC:  Homotypic versus heterotypic continuity of anxiety symptoms in young adolescents: evidence for distinctions between DSM-IV subtypes.  J Abnorm Child Psychol 2007; 35:325—333
[CrossRef] | [PubMed]
 
Kendler  KS;  Gardner  CO;  Lichtenstein  P:  A developmental twin study of symptoms of anxiety and depression: evidence for genetic innovation and attenuation.  Psychol Med 2008; 38:1567—1575
[CrossRef] | [PubMed]
 
Constantino  J;  Todd  RD:  Autistic traits in the general population: a twin study.  Arch Gen Psychiatry 2003; 60:524—530
[CrossRef] | [PubMed]
 
Haberstick  BC;  Schmitz  S;  Young  SE;  Hewitt  JK:  Contributions of genes and environments to stability and change in externalizing and internalizing problems during elementary and middle school.  Behav Genet 2005; 35:381—396
[CrossRef] | [PubMed]
 
Weisbrot  DM;  Gadow  KD;  DeVincent  CJ;  Pomeroy  J:  The presentation of anxiety in children with pervasive developmental disorders.  J Child Adolesc Psychopharmacol 2005; 15:477—496
[CrossRef] | [PubMed]
 
References Container
+
+

Self-Assessment Quiz - Expired

Did you know? You can add a subscription now to earn CME Credits!

1.
The association between autistic-like and internalizing traits was compared between middle and late childhood in twin pairs. What pattern was seen in the intraclass twin correlations?
2.
In the analyses of the influence of autistic-like traits on later internalizing traits, which of the following showed the strongest association with internalizing traits at the second timepoint?
3.
With regard to autistic-like traits scores, which of the following statements best describes the pattern for boys relative to girls?
Submit a Comments
Please read the other comments before you post yours. Contributors must reveal any conflict of interest.
Comments are moderated and will appear on the site at the discertion of APA editorial staff.

* = Required Field
(if multiple authors, separate names by comma)
Example: John Doe



Web of Science® Times Cited: 17

Related Content
Articles
Books
The American Psychiatric Publishing Textbook of Psychopharmacology, 4th Edition > Chapter 3.  >
The American Psychiatric Publishing Textbook of Psychopharmacology, 4th Edition > Chapter 3.  >
Dulcan's Textbook of Child and Adolescent Psychiatry > Chapter 13.  >
Dulcan's Textbook of Child and Adolescent Psychiatry > Chapter 23.  >
Dulcan's Textbook of Child and Adolescent Psychiatry > Chapter 16.  >
Topic Collections
Psychiatric News
Read more at Psychiatric News >>
PubMed Articles