Introduction
Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by social impairments and restricted and repetitive behaviors or interests (Hirota & King, Reference Hirota and King2023). The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) described that ASD symptoms are dimensional and represent an extreme end of autistic-like traits, in accordance with the neurodiversity perspective (American Psychiatric Association, 2013; J. N. Constantino & Todd, Reference Constantino and Todd2003; Gillberg, Reference Gillberg1992). This dimension, acknowledged as autistic-like traits, refers to behavioral patterns that resemble characteristics commonly associated with ASD and on a continuum in the general population, typically including differences in social communication (e.g. interest in social interaction) (J.N. Constantino & Gruber, Reference Constantino and Gruber2005; J. N. Constantino & Todd, Reference Constantino and Todd2003; Posserud, Lundervold, & Gillberg, Reference Posserud, Lundervold and Gillberg2006). The social communication abilities demonstrate substantial heritability (approximately 50%), comparable to ASD (J. N. Constantino & Todd, Reference Constantino and Todd2003). Meanwhile, these traits are also modulated by environmental factors, including family dynamics (Gerstein & Crnic, Reference Gerstein and Crnic2018; Martini et al., Reference Martini, Butwicka, Du Rietz, Kanina, Rosenqvist, Larsson and Taylor2024). One longitudinal study has indicated that subclinical autistic social and communication traits may vary in their severity from childhood to adolescence (Pender, Fearon, Pourcain, Heron, & Mandy, Reference Pender, Fearon, Pourcain, Heron and Mandy2023). Specifically, there were 7.3% of children undergoing a dramatic increase in autistic-like traits from 10 to 16 years old, and another 6.9% of children experienced a decrease in autistic-like traits over adolescence. The temporal variability might be contributed to environmental and genetic triggers starting at puberty, and non-autistic social communication difficulties secondary to depression and anxiety in adolescence. Elucidating the environmental risk factors associated with autistic-like traits during this critical developmental period would enhance our understanding of their etiology. However, empirical evidence is lacking on this topic.
Parents of children with ASD frequently experience psychopathological problems, such as depression, anxiety, interpersonal sensitivity, paranoid ideation, obsessive–compulsive behaviors, and somatic complaints (Catalano, Holloway, & Mpofu, Reference Catalano, Holloway and Mpofu2018; Hodge, Hoffman, & Sweeney, Reference Hodge, Hoffman and Sweeney2011). Also, maternal depressive symptoms, anxiety, obsessive–compulsive symptoms, emotional problems, and symptoms of attention-deficit/hyperactivity disorder (ADHD) during and after pregnancy were linked to more social communication problems and peer problems in childhood and adolescence, though the effect sizes were generally small to moderate (Amiri et al., Reference Amiri, Lamballais, Geenjaar, Blanken, El Marroun, Tiemeier and White2020; Efron, Furley, Gulenc, & Sciberras, Reference Efron, Furley, Gulenc and Sciberras2018; El Marroun et al., Reference El Marroun, White, van der Knaap, Homberg, Fernández, Schoemaker and Tiemeier2014; Goh et al., Reference Goh, Gan, Kung, Baron-Cohen, Allison, Chen and Magiati2018; Kleine et al., Reference Kleine, Falconer, Roth, Counsell, Redshaw, Kennea and Nosarti2020). Parental psychopathology has been identified as a risk factor for family conflict, which may in turn affect child autistic-like traits (Agha, Zammit, Thapar, & Langley, Reference Agha, Zammit, Thapar and Langley2013; Zhang, Lee, White, & Qiu, Reference Zhang, Lee, White and Qiu2020). Moreover, brain functions undergo significant developmental changes during early adolescence, with functional brain connectivity exhibiting high plasticity (Dahl, Reference Dahl2004). Previous studies have suggested neurobiological dysfunction as a potential mechanism of intergenerational transmission of psychopathologies such as depression and ADHD (Beardslee, Gladstone, & O’Connor, Reference Beardslee, Gladstone and O’Connor2011; Epstein et al., Reference Epstein, Casey, Tonev, Davidson, Reiss, Garrett and Spicer2007). Adverse family environments have been recognized to influence children’s behavioral problems by modifying specific functional brain connectivity patterns (Zhi et al., Reference Zhi, Jiang, Pearlson, Fu, Qi, Yan and Sui2024), which also underlie autistic-like traits (Nenadić et al., Reference Nenadić, Schröder, Hoffmann, Evermann, Pfarr, Bergmann and Meller2024; Tu et al., Reference Tu, Hsu, Lan, Liu, Su and Chen2016). However, most existing studies examined isolated dimensions of parental psychopathology rather than employing a comprehensive assessment (Loeber, Hipwell, Battista, Sembower, & Stouthamer-Loeber, Reference Loeber, Hipwell, Battista, Sembower and Stouthamer-Loeber2009). Given the well-documented high correlations between various psychopathologies (Guo et al., Reference Guo, Cui, Qiu, Bu, Yang, Li and Zhu2024; Hartman et al., Reference Hartman, Larsson, Vos, Bellato, Libutzki, Solberg and Reif2023), the observed associations may be confounded by unmeasured comorbid conditions. Additionally, previous studies were mostly cross-sectional, limiting the ability to examine the directionality of this association.
This study aims to fill these research gaps using a large cohort of adolescents in the United States. First, we hypothesized that specific parental psychopathology could influence child autistic-like traits in early adolescence. We applied a network approach to identify the most important parental psychopathology linked to child autistic-like traits. We then implemented two-wave cross-lagged panel models (CLPMs) to investigate the directionality of longitudinal associations. Second, we hypothesized that at the behavioral level, family conflict might mediate the association between parental psychopathology and autistic-like traits; at the neurobiological level, brain functional connectivity might mediate the association. Mediation analyses were applied. The findings are expected to inform clinical interventions aimed at promoting both parental and adolescent mental health.
Methods
Participants
Our study was comprised of participants from the Adolescent Brain Cognitive Development (ABCD) study (release 5.0; https://abcdstudy.org/), which recruited over 11,000 children aged 9–11 years from 21 centers across the United States (Garavan et al., Reference Garavan, Bartsch, Conway, Decastro, Goldstein, Heeringa and Zahs2018). Parents’ written informed consent and children’s assent were obtained at recruitment (Clark et al., Reference Clark, Fisher, Bookheimer, Brown, Evans, Hopfer and Yurgelun-Todd2018). Of the children with complete parental psychopathology and child autistic-like traits information (N = 10,313), we randomly sampled one child from each family to account for the family-clustered structure (N = 8,571).
Measures
Parental psychopathology was assessed using the Adult Self Report (ASR) from the Achenbach System of Empirically Based Assessment at baseline and 2-year follow-up (Achenbach, Reference Achenbach2009). The ASR consists of 120 items and scores on six DSM-oriented dimensions: depressive, anxiety, somatic, avoidant personality, attention-deficit/hyperactivity, and antisocial personality problems. In addition, we computed 8 dimension-based composite scores of the syndrome scales utilizing an alternative established algorithm (i.e. anxious/depressed, withdrawn, somatic complaints, thought problems, attention problems, aggressive behavior, rule-breaking behavior, and Intrusive) (Achenbach, Bernstein, & Dumenci, Reference Achenbach, Bernstein and Dumenci2005). Parents rated each item on a 3-point Likert scale ranging from 0 (not true) to 2 (very true or often true), reflecting their experiences in the past 6 months. The ASR has shown good reliability and validity (Achenbach & Rescorla, Reference Achenbach and Rescorla2003). We computed composite scores for each dimension of psychopathology, with higher scores reflecting greater severity of psychopathology.
Previous studies have reported different measures of autistic-like traits, including: the Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5 (KSADS-5, ASD subscale), Child Behavior Checklist (CBCL, ASD subscale), and Social Responsiveness Scale (SRS), respectively. In the ABCD study, these three measures showed acceptable validity for identifying parent-reported ASD (area under the curve, 0.80 for KSADS-5, 0.81 for CBCL, and 0.90 for SRS). Meanwhile, the correlations between these measures were weak to moderate (Spearman’s ρ, 0.38–0.52; Table S1). Thus, we performed analyses on all three measures for comparison.
The KSADS-5, a widely-used screening tool for mental disorders, contains three parent-reported and autism-related questions (i.e. unusual body movements, strict routines, and poor eye contact during the past 2 weeks) which showed acceptable reliability and validity for screening ASD (Barch et al., Reference Barch, Albaugh, Avenevoli, Chang, Clark, Glantz and Sher2018; Townsend et al., Reference Townsend, Kobak, Kearney, Milham, Andreotti, Escalera and Kaufman2020; Ünal et al., Reference Ünal, Öktem, Çetin Çuhadaroğlu, Çengel Kültür, Akdemir, Foto Özdemir and Artık2019). Each item score ranges from 0 (not at all) to 4 (nearly every day); the total score ranges from 0 to 12. The CBCL contains 113 items that measure emotional and behavioral problems (Achenbach, Reference Achenbach2009), 15 of which were used to measure child ASD-related behavioral and emotional problems across the past 6 months (Offermans et al., Reference Offermans, de Bruin, Lange, Middeldorp, Wesseldijk, Boomsma and van Steensel2023). Parents rated each item on a 3-point Likert scale ranging from 0 (not true) to 2 (very true or often true), and the total score ranges from 0 to 30. Parents also completed the 11-item abridged version of SRS, which was also applied in screening for ASD (Reiersen, Constantino, Grimmer, Martin, & Todd, Reference Reiersen, Constantino, Grimmer, Martin and Todd2008). Each item was rated using a 4-point Likert scale ranging from 0 (not true) to 3 (almost always true); the total score ranges from 0 to 33.
For all three measures, higher scores represent more evident autistic-like traits. To be noted, the KSADS-5 and CBCL data were available at both baseline and 2-year follow-up, while the SRS was only available at 1-year follow-up.
Covariates
Potential confounders included child age, sex (male, or female), race/ethnicity (white, black, Hispanic, or others), family annual income level ($34,999 and less, $35,000 through $74,999, $75,000 through $99,999, or $100,000 and higher), and recruitment site. The highest missing proportion was 8.1% for the family annual income level. Missing data were imputed using the multivariate imputation by chained equations (van Buuren & Groothuis-Oudshoorn, Reference van Buuren and Groothuis-Oudshoorn2011). Previous studies suggested that autistic-like traits and behavioral problems are distinct but correlated constructs (Lundström et al., Reference Lundström, Chang, Kerekes, Gumpert, Råstam, Gillberg and Anckarsäter2011). Thus, we additionally adjusted for the CBCL internalizing and externalizing problem scores of children at baseline to specifically investigate autistic-like traits. ASD and autistic-like traits have high heritability (Serdarevic et al., Reference Serdarevic, Tiemeier, Jansen, Alemany, Xerxa, Neumann and Ghassabian2020). To investigate to what extent the observed associations could be explained by genetics, in an additional analysis, we further adjusted for the polygenic risk score (PRS) for ASD. We were also interested in the mediating role of family conflict and resting-state functional magnetic resonance imaging (fMRI) connectivity in the observed associations between parental psychopathology and child autistic-like traits.
PRS for ASD
The genome-wide association study of autism was from the latest release of the iPSYCH cohort, involving 18,382 ASD and 27,969 controls of European ancestry (Grove et al., Reference Grove, Ripke, Als, Mattheisen, Walters, Won and Børglum2019). PRS scores were generated for ASD using polygenic risk scoring with continuous shrinkage (Ge, Chen, Ni, Feng, & Smoller, Reference Ge, Chen, Ni, Feng and Smoller2019). Quality control was performed by PLINK v1.90, and missing data was imputed using the Michigan Imputation Server, 40 with the 1,000 Genomes Project EUR (Phase 3, hg19) reference panel and Eagle v2.4 phasing (Warrier et al., Reference Warrier, Zhang, Reed, Havdahl, Moore, Cliquet and Baron-Cohen2022; Yang et al., Reference Yang, Rolls, Dong, Du, Li, Feng and Zhao2022). Finally, 3,548 white individuals with complete parental psychopathology and child autistic-like traits data were included in the analyses involving PRS.
Family conflict
Family conflict was measured by the Family Conflict subscale of the Family Environment Scale (FES) at baseline (Moos & Moos, Reference Moos and Moos1994). This is a 9-item dichotomous questionnaire reported by parents (Zucker et al., Reference Zucker, Gonzalez, Feldstein Ewing, Paulus, Arroyo, Fuligni and Wills2018). The scale ranges from 0 to 9, and higher scores indicate a more conflictual family environment.
Resting-state fMRI connectivity
Imaging acquisition, scanning parameters, and preprocessing procedures are described in detail by the ABCD team (Casey et al., Reference Casey, Cannonier, Conley, Cohen, Barch, Heitzeg and Dale2018; Hagler et al., Reference Hagler, Hatton, Cornejo, Makowski, Fair, Dick and Dale2019). fMRI data were collected across sites with harmonized protocols on 3T scanner platforms, including Siemens Magnetom Prisma, General Electric Discovery MR750, and Philips Achieva scanners. Participants completed four 5-minute resting-state scans with eyes open to ensure at least 8 minutes of relatively low-motion data. Intra- and inter-network-level resting-state functional connectivity (rsFC; Pearson correlation) were calculated based on the Gordon parcellation scheme to group cortical-surface regions into 12 predefined resting-state networks (Gordon et al., Reference Gordon, Laumann, Adeyemo, Huckins, Kelley and Petersen2016). Of these 12 networks, we specifically examined 8 networks that were reported to be associated with ASD, including auditory network (AN), cingulo-opercular network (CON), default mode network (DMN), dorsal attention network (DAN), fronto-parietal network (FPN), salience network (SN), ventral attention network (VAN), and visual network (VN) (Bi, Zhao, Xu, Sun, & Wang, Reference Bi, Zhao, Xu, Sun and Wang2018; Shan et al., Reference Shan, Uddin, Ma, Xu, Xiao, Li and Duan2023; Sun et al., Reference Sun, Fan, Wang, Wang, Jia and Ma2021). They were Fischer z transformed, resulting in 36 network-level rsFC correlation averages (8 intra- and 28 inter-circuits).
Statistical analysis
All analyses were performed using R version 4.3.1.
Cross-sectional network analyses
A Graphical Gaussian Model (GGM) of network analyses provides a novel approach to visualizing the complex interaction among multiple symptoms based on correlations between each pair of symptoms (Epskamp, Waldorp, Mõttus, & Borsboom, Reference Epskamp, Waldorp, Mõttus and Borsboom2018). We estimated GGMs to investigate the relationship among parental psychopathologies (ASR at baseline) and child autistic-like traits (KSADS-5 and CBCL at baseline, and SRS at 1-year follow-up). In this graph, each node corresponds to a symptom (parental dimensional psychopathologies and child autistic-like traits), and the edge weight indicates the strength of the partial correlation between the two symptoms. An optimal unregularized GGM with the lowest Extended Bayesian Information Criterion was selected for further analyses, which balanced model parsimony (number of edges) against model fit. Edges contributing minimally to log-likelihood were eliminated (Chen & Chen, Reference Chen and Chen2008; Foygel & Drton, Reference Foygel and Drton2010). We applied flow diagrams with the package ‘qgraph’ (version 1.9.5) to visualize the network structure (Epskamp, Cramer, Waldorp, Schmittmann, & Borsboom, Reference Epskamp, Cramer, Waldorp, Schmittmann and Borsboom2012). This study mainly focused on edge weights between child autistic-like traits and six parental dimensional psychopathologies. We estimated node centrality, which measured the relative importance of a node within a network, including strength, closeness, betweenness, and expected influence (definitions detailed in Table S2) (McNally, Reference McNally2016). Network analyses were also performed using the 8 dimension-based composite scores from the syndrome scales, for comparison.
The network stability was examined using bootstrap methods with the package ‘bootnet’ (version 1.5.3) in following steps: (a) implementing case-dropping bootstrap to investigate the stability of centrality indices; (b) implementing nonparametric bootstrap to draw confidence intervals (CIs) of edge weights and test for significant differences between edge weights and centrality indices (Epskamp, Borsboom, & Fried, Reference Epskamp, Borsboom and Fried2018; Martin, Maughan, Konac, & Barker, Reference Martin, Maughan, Konac and Barker2023). The correlation stability coefficient refers to the maximum proportion of cases that can be dropped, such that with 95% certainty, the correlation between original centrality indices and centrality of networks based on subsets is at least 0.7. It was applied to quantify the case-dropping bootstrap and should not be below 0.25.
We investigated the child sex difference of the network by permutation-based tests to examine differences between networks based on global strength invariance (i.e. the absolute sum of edge weights) and network structure invariance (i.e. distributions of edges) (van Borkulo et al., Reference van Borkulo, van Bork, Boschloo, Kossakowski, Tio, Schoevers and Waldorp2023). The analysis was performed using the package ‘NetworkComparisonTest’ (version 2.2.1). We performed the following sensitivity analyses to test the robustness of the network. First, we added children’s behavioral problems to the network model and identified specific parental psychopathology that directly correlated with child autistic-like traits. Second, we re-estimated network models, additionally involving potential confounders in the network nodes, including child age, sex, race/ethnicity, family annual income level, and recruitment site. To be noted, adding children’s behavioral problems to the network led to the elimination of almost all other edge weights in the unregularized GGM, due to the strong correlations between behavioral problems and autistic-like traits. Thirdly, we identified the eldest child from each family (N = 8,571) and re-estimated network models for comparison.
Longitudinal data analyses
Two-wave CLPMs based on structural equation modelling were implemented to investigate the longitudinal association between parental psychopathology and child autistic-like traits (KSADS-5 and CBCL) from baseline to 2-year follow-up with package ‘lavaan’ (version 0.6.15) (Shen et al., Reference Shen, Luo, Chamberlain, Morgan, Romero-Garcia, Du and Sahakian2020). We adjusted for child age, sex, race/ethnicity, family annual income level, internalizing symptom score, externalizing symptom score, and recruitment site. A comparative fit index (CFI) >0.90 or a standardized root mean square residual (SRMR) <0.08 suggests adequate model fit (Hu & Bentler, Reference Hu and Bentler1999). A false discovery rate (FDR) correction using the Benjamini–Hochberg method was conducted for all 12 cross-lagged paths (6 parental dimensional psychopathologies × 2 directions). An FDR < 0.05 indicated statistical significance. We performed sensitivity analyses by the regression-with-residuals method to adjust for child internalizing and externalizing symptom scores at both baseline and 2-year follow-up, along with all other covariates from the main analysis. We also investigated the potential confounding role of genetic predisposition to ASD by additionally adjusting for PRS for ASD in a subsample of white children (N = 3,548).
Mediation analyses
As the longitudinal association of parental attention-deficit/hyperactivity problems with child autistic-like traits was determined by CLPMs, we further examined the mediation effect of family conflict and functional brain connectivity on this association at baseline (Kerr-German, White, Santosa, Buss, & Doucet, Reference Kerr-German, White, Santosa, Buss and Doucet2022; Lau et al., Reference Lau, Hawes, Hunt, Frankland, Roberts, Wright and Mitchell2018; Modabbernia, Janiri, Doucet, Reichenberg, & Frangou, Reference Modabbernia, Janiri, Doucet, Reichenberg and Frangou2021; Zhang et al., Reference Zhang, Lee, White and Qiu2020). The mediation analyses applied a standard 3-variable path model with adjustment for child age, sex, race/ethnicity, family annual income level, and recruitment site (Baron & Kenny, Reference Baron and Kenny1986). The significance of the mediation was estimated by the bias-corrected bootstrap approach (with 1,000 random samplings). In this analysis aiming to investigate the potential mechanisms, we could not rule out that behavioral problems lay on the pathway between parental psychopathology and child autistic-like traits. Thus, in the main analyses, we did not adjust for behavioral problems to avoid overadjustment. In an additional analysis, we further adjusted for externalizing and internalizing symptom scores for comparison.
Results
Table 1 shows the demographic characteristics of the children involved in this study (N = 8,571; mean age, 9.5 [SD, 0.5] years; 53% boys; 53% white; 42.5% with family annual income > $100,000).
Demographic characteristics of the eligible children from the Adolescent Brain Cognitive Development study
Abbreviations: KSADS-5, Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5; CBCL, Child Behavior Checklist; SRS, Social Responsiveness Scale; SD, standard deviation.
a Parental psychopathology was measured by self-reported Adult Self Report.
b Family conflict was measured by parent-reported Family Conflict subscale of Family Environment Scale at baseline.
c Child co-occurring behavioral problem was measured by parent-reported Child Behavior Checklist at baseline.
d These three scales were all parent-reported.
Cross-sectional network analyses
Figure 1 presents the network structure between parental dimensional psychopathology and child autistic-like traits, measured by three different methods. Each network had a network density of 0.9 and a mean edge weight of 0.1. Parental attention-deficit/hyperactivity problems had a direct and the strongest connection with child autistic-like traits, irrespective of the measurement methods of the traits. Within each network, the centrality values of parental attention-deficit/hyperactivity problems ranked second, following parental depressive problems. Meanwhile, the connection between parental depressive problems and child autistic-like traits was weak, irrespective of the measurement methods (ranked 5th to 6th out of 6 dimensional psychopathologies). Additional network analyses employing an 8-factor structure of the ASR scales revealed inconsistent results in the analyses utilizing different measures of autistic-like traits (Figure S1). The inconsistency limits further interpretation and investigation utilizing this approach.
Network analyses of parental psychopathology and child autistic-like traits. Child autistic-like traits were assessed by (a–c) KSADS-5, (d–f) CBCL, and (g–i) SRS, respectively. (a, d, g) Network structure, (b, e, h) edge weight ranks between child autistic-like traits and six parental dimensional psychopathologies, and (c, f, i) node centrality measures are presented, respectively. Abbreviations: KSADS-5, Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5; CBCL, Child Behavior Checklist; SRS, Social Responsiveness Scale; ALTs, autistic-like traits; AD/H, attention-deficit/hyperactivity.
To test the robustness of the network, we observed that the centrality indices were stable in each network (correlation stability coefficient = 0.75, indicating that 75% of the sample could be dropped while the network structure did not change significantly; Figure S2). The nonparametric bootstrap tests revealed that the edge weights and centrality indices differed significantly for most comparisons (Figures S3 andS4).
The distribution plots of network comparison tests (Figure S5) illustrated that the global strength and network structure did not show significant sex differences in the three networks (e.g. KSADS-5-measured: strengthdiff < 0.001, p = 0.998; edgemaxdiff = 0.065, p = 0.125). In the network models involving child internalizing and externalizing symptoms (Figure S6), parental attention-deficit/hyperactivity problem was the only parental psychopathology that showed consistent associations with children’s autistic-like traits across multiple measures (CBCL and SRS, though not observed using the KSADS-5). Additionally, adding potential confounders into network models did not essentially alter the pattern and magnitude of associations (Tables S3 and S4). The sensitivity analyses using the eldest participant in each family of the ABCD study showed consistent results (Figure S7).
Longitudinal data analyses
We investigated longitudinal associations between parental psychopathology and child autistic-like traits at baseline and 2-year follow-up. After adjusting for demographic variables and internalizing and externalizing problems, we observed that the only significant cross-lagged association was between parental attention-deficit/hyperactivity problems at baseline and child autistic-like traits at 2-year follow-up, regardless of measurement methods (KSADS-5-measured: β = 0.014, 95% CI [0.010, 0.018], FDR q = 0.005; CBCL-measured: β = 0.060, 95% CI [0.052, 0.068], FDR q < 0.001; Figure 2 and Table S5). Additionally, adjusting for child internalizing and externalizing symptom scores at 2-year follow-up revealed similar results for CBCL-measured autistic-like traits, while the KSADS-5 measured result was not statistically significant (Table S6).
Two-wave cross-lagged panel models of parental attention-deficit/hyperactivity problems and child autistic-like traits. Child autistic-like traits were assessed by (a) KSADS-5 and (b) CBCL, respectively. Abbreviations: KSADS-5, Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5; CBCL, Child Behavior Checklist; SE, standard error. Note: Solid lines represent statistical significance (p < 0.05), and dashed lines represent non-significance (p > 0.05).
To account for the potential genetic factor underlying the observed associations, we calculated PRS for ASD in a subsample of white children (N = 3,548). Additionally adjusting for PRS for ASD did not essentially alter the effect size in CLPMs (Table S7 in the Supplementary Material).
Mediation analyses
We examined the potential mediating role of family conflict and functional brain connectivity in the observed association between parental attention-deficit/hyperactivity problems and child autistic-like traits. Parental attention-deficit/hyperactivity problems were associated with a higher family conflict score (β = 0.165, 95% CI [0.159, 0.171], p < 0.001). Using three different measures for child autistic-like traits, we robustly observed that family conflict mediated the association between parental attention-deficit/hyperactivity problems and child autistic-like traits (KSADS-5-measured: proportion mediated = 11.5%, p for indirect effect <0.001; CBCL-measured: proportion mediated = 8.0%, p for indirect effect <0.001; SRS-measured: proportion mediated = 6.8%, p for indirect effect <0.001; Figure 3a and Figure S8A,B).
Mediation analyses of (a) family conflict and (b) average correlation between default mode network and dorsal attention network in the association between parental attention-deficit/hyperactivity problems and child autistic-like traits, measured by KSADS-5. Abbreviations: KSADS-5, Kiddie Schedule for Affective Disorders and Schizophrenia for DSM-5; SE, standard error.
We observed that parental attention-deficit/hyperactivity problems were associated with an increased DMN-DAN connectivity in children (β = 0.011, 95% CI [0.008, 0.114], p < 0.001, FDR q = 0.029; Table S8). DMN-DAN connectivity in children significantly mediated the observed association between parental attention-deficit/hyperactivity problems and child autistic-like traits (KSADS-5-measured: proportion mediated = 0.7%, p for indirect effect = 0.047; CBCL-measured: proportion mediated = 0.4%, p for indirect effect = 0.025; Figure 3b and Figure S8C). In the analysis on the SRS-measured autistic-like traits, we did not observe a statistically significant indirect effect of the DMN-DAN connectivity (p for indirect effect = 0.207; Figure S8D). In the additional analyses, further adjusted for externalizing and internalizing problems, the indirect effect of family conflict and DMN-DAN connectivity was not statistically significant (Figure S9).
Discussion
In a large cohort from the ABCD study, we first investigated how parental psychopathology influences autistic-like traits in early adolescents, then explored potential mediating mechanisms. We used a network approach and highlighted that parental attention-deficit/hyperactivity problems had the strongest link among psychopathologies that were directly associated with child autistic-like traits in early adolescence. In longitudinal analyses, we observed that parental attention-deficit/hyperactivity problems at baseline showed a robust link to child autistic-like traits at follow-up. Family conflict and DMN-DAN connectivity significantly mediate the association between parental attention-deficit/hyperactivity problems and child autistic-like traits.
Previous studies have reported that specific parental dimensional psychopathology was associated with child autistic-like traits. Goh et al. (Reference Goh, Gan, Kung, Baron-Cohen, Allison, Chen and Magiati2018) found that maternal depressive symptoms at 24 months were positively associated with toddler social-communicative autistic-like traits. Amiri et al. (Reference Amiri, Lamballais, Geenjaar, Blanken, El Marroun, Tiemeier and White2020) studied 3,942 children and found that maternal anxiety, obsessive–compulsive symptoms, and difficulties concentrating during prenatal and perinatal life were associated with later development of child autistic-like traits. Our network analyses, accounting for the high comorbidity of psychopathologies, corroborated previous findings and highlighted that parental attention-deficit/hyperactivity problems had a direct and the strongest link with child autistic-like traits. Importantly, in the longitudinal analyses accounting for the correlation between behavioral problems and autistic-like traits (Lundström et al., Reference Lundström, Chang, Kerekes, Gumpert, Råstam, Gillberg and Anckarsäter2011; Simonoff et al., Reference Simonoff, Pickles, Charman, Chandler, Loucas and Baird2008), the only statistically significant cross-lagged association was between parental attention-deficit/hyperactivity problems at baseline and child autistic-like traits at 2-year follow-up. Our findings extend the current knowledge that parental attention-deficit/hyperactivity problems could be a risk factor for child elevated autistic-like traits during early adolescence.
Multiple mechanisms have been identified concerning parent–child psychopathology transmission, including dysfunctional family environment, neurobiological dysfunction, and genetic predisposition (Beardslee et al., Reference Beardslee, Gladstone and O’Connor2011; Ferreira et al., Reference Ferreira, Moreira, Kleinman, Nader, Gomes, Teixeira and Caetano2013; Gebru et al., Reference Gebru, Goncalves, Cruz, Thompson, Allegair, Potter and Johnson2023). Parental attention-deficit/hyperactivity symptoms have been identified as a risk factor for child social problems (Efron et al., Reference Efron, Furley, Gulenc and Sciberras2018; Zhang et al., Reference Zhang, Lee, White and Qiu2020). Compared with other parental psychopathological dimensions (Jones, Hall, & Kiel, Reference Jones, Hall and Kiel2021; Lovejoy, Graczyk, O’Hare, & Neuman, Reference Lovejoy, Graczyk, O’Hare and Neuman2000), parental attention-deficit/hyperactivity problems were more prominently associated with unfavorable parenting, such as hostility, criticism, physical or verbal aggression, inconsistent discipline, indulgent, neglectful, and lack of sensitivity and responsiveness to the child (Park, Hudec, & Johnston, Reference Park, Hudec and Johnston2017), which may in turn arouse family conflict and affect child social behavior (Agha et al., Reference Agha, Zammit, Thapar and Langley2013). Consistently, we observed that family conflict significantly mediated the association between parental attention-deficit/hyperactivity problems and child autistic-like traits.
ASD and ADHD exhibited shared aberrant functional brain connectivity, other than depression and anxiety (Di Martino et al., Reference Di Martino, Zuo, Kelly, Grzadzinski, Mennes, Schvarcz and Milham2013; Lukito et al., Reference Lukito, Norman, Carlisi, Radua, Hart, Simonoff and Rubia2020). Specifically, DMN is an integrated system typically activated during self-related cognitive processes, including autobiographical memory, internal thoughts, and social functions (Menon, Reference Menon2011), while DAN involves goal-directed top-down orienting of attention (Fox, Corbetta, Snyder, Vincent, & Raichle, Reference Fox, Corbetta, Snyder, Vincent and Raichle2006). The anticorrelation of these two networks is considered healthy functional brain connectivity (Fox et al., Reference Fox, Snyder, Vincent, Corbetta, Van Essen and Raichle2005), and this altered anticorrelation has been found to relate to ASD and ADHD (McKinnon et al., Reference McKinnon, Eggebrecht, Todorov, Wolff, Elison, Adams and Pruett2019; Strang et al., Reference Strang, McClellan, Li, Jack, Wallace, McQuaid and Vaidya2023; Tomasi & Volkow, Reference Tomasi and Volkow2012). Our findings were consistent with previous studies that supported the relationship between DMN-DAN connectivity and autistic-like traits (McKinnon et al., Reference McKinnon, Eggebrecht, Todorov, Wolff, Elison, Adams and Pruett2019; Strang et al., Reference Strang, McClellan, Li, Jack, Wallace, McQuaid and Vaidya2023), and our further mediation analysis showed that increased DMN-DAN connectivity in resting-state fMRI significantly mediated the association between parental attention-deficit/hyperactivity problems and child autistic-like traits. Interestingly, previous studies reported a generally decreased DMN-DAN connectivity as shared neurobiological dysfunction of ASD and ADHD in children at adolescence (Kernbach et al., Reference Kernbach, Satterthwaite, Bassett, Smallwood, Margulies, Krall and Bzdok2018). This discrepancy in direction could be explained by the overall developmental shift from hyper-connectivity in childhood to hypo-connectivity in adolescence and adulthood (Padmanabhan, Lynch, Schaer, & Menon, Reference Padmanabhan, Lynch, Schaer and Menon2017), considering that the children in this analysis were at a mean age of 9.5 years and at the transitional stage from childhood to adolescence. In addition, we observed that adjusting for behavioral problems would nullify the indirect effect through family conflict and DMN-DAN connectivity, indicating a common etiology of behavioral problems and autistic-like traits related to parental attention-deficit/hyperactivity problems.
ADHD is among the most common psychiatric comorbidities in ASD (28% prevalence, compared to ~20% for anxiety and ~10% for depression) (Hirota & King, Reference Hirota and King2023), with substantial genetic risk overlap between the conditions (Stergiakouli et al., Reference Stergiakouli, Davey Smith, Martin, Skuse, Viechtbauer, Ring and St2017; Wu et al., Reference Wu, Cao, Baranova, Huang, Li, Cai and Wang2020). Considering the shared genetic predisposition of ADHD and ASD, we additionally adjusted for PRS for ASD in a subsample of white children, observing comparable risk estimates with the main analysis. Though this may indicate a limited role of genetic predisposition, it should also be noted that PRS for ASD could only explain less than 5% of the etiology of ASD, despite ASD’s high heritability (>50%) (Grove et al., Reference Grove, Ripke, Als, Mattheisen, Walters, Won and Børglum2019). Therefore, we cannot rule out that shared genetic factors between ADHD and ASD may still partly explain the observed association between parental attention-deficit/hyperactivity problems and children’s autistic-like traits.
Parental depression has been identified as a risk factor for child autistic-like traits (Avalos et al., Reference Avalos, Chandran, Churchill, Gao, Ames, Nozadi and Croen2023; Breider et al., Reference Breider, Hoekstra, Wardenaar, van den Hoofdakker, Dietrich and de Bildt2022; Goh et al., Reference Goh, Gan, Kung, Baron-Cohen, Allison, Chen and Magiati2018), and often coexists with other psychopathologies (Daga et al., Reference Daga, Gramaglia, Bailer, Bergese, Marzola and Fassino2011; Sandstrom, Perroud, Alda, Uher, & Pavlova, Reference Sandstrom, Perroud, Alda, Uher and Pavlova2021; Tiller, Reference Tiller2013). Our network models revealed that parental depressive problems had the highest centrality index, while in two of the three networks, parental depressive problems had the lowest association with child autistic-like traits. As centrality index indicates relative importance of a node within a network (Robinaugh, Millner, & McNally, Reference Robinaugh, Millner and McNally2016), our results implied that parental depressive problems were most strongly related to other dimensional psychopathological symptoms, while they connected to offspring autistic traits mainly through other psychopathological symptoms.
The study has the following strengths. First, this study included a large sample with repeated measures data that permitted both cross-sectional network and longitudinal analyses. By combining the two methods, we provided evidence that may suggest the directionality of observed associations. Second, as the measure of autistic-like traits remains controversial, we used three measures of child autistic-like traits, and the results were consistent, indicating high validity of the findings. Third, the genetic, familial, and functional brain connectivity data enabled us to explore the possible effect-modifiers or mediators of the observed associations.
The study should be interpreted in the light of the following limitations. First, we lacked data on whether the psychopathology belonged to the mother or the father. A previous study suggested that the impact of maternal psychopathology on child development may differ from that of paternal psychopathology (Rogers et al., Reference Rogers, Youssef, Teague, Sunderland, Le Bas, Macdonald and Hutchinson2023). Second, the exposure and outcomes in this study were exclusively parent-reported, which may introduce measurement bias and inflate the strength of observed associations. While we attempted to mitigate this limitation by incorporating statistical adjustments for children’s behavioral problems, the fundamental challenge of disentangling parental perceptions – potentially influenced by their own psychopathological traits – from children’s actual behavioral manifestations remains. This methodological constraint underscores the need for future investigations to incorporate direct, objective assessments of autistic-like traits in children, such as tests of theory of mind, executive functions, and pragmatic language. Third, though we utilized three established instruments to assess autistic-like traits, studies should employ more comprehensive measurements, including Childhood Autism Rating Scale-Second Edition Questionnaire of Parent Concerns, the Autism Treatment Evaluation Checklist, Mental Synthesis Evaluation Checklist, and Behavior Rating Inventory of Executive Function. Fourth, the follow-up period was relatively short. Future research with extended follow-up data is required to better comprehend the variations in autistic traits in the crucial period from childhood to adolescence.
Conclusion
In summary, we observed a robust association between parental attention-deficit/hyperactivity problems at baseline and child autistic-like traits at follow-up during early adolescence. Family conflict and dysfunctional DMN-DAN connectivity significantly mediated the association.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725100779.
Data availability statement
The data that support the findings of this study are openly available in the ABCD Dataset Data Release 5.0 at https://nda.nih.gov/abcd. The scripts used for these analyses will be made available upon publication (doi: 10.15154/dkxz-0k07).
Acknowledgements
Data used in the preparation of this article were obtained from the ABCD study (https://abcdstudy. org), held in the NIMH Data Archive. This is a multisite, longitudinal study designed to recruit more than 10,000 children aged 9–10 years and follow them over 10 years into early adulthood. The ABCD study is supported by the National Institutes of Health and additional federal partners under award numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, U01DA041148, U01DA041093, and U01DA041089. A full list of supporters is available at https://abcdstudy.org/federal-partners.html. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/scientists/workgroups/. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators.
Author contributions
MW, YQL, TLZ, QL, TR and FL designed the study. MW and TR conducted data analysis. MW, YQL, TLZ, and TR drafted the manuscript. MW, YQL, TLZ, RQH, LKH, LLZ, QLZ, YJS, WZ, YWP, JYC, HH, SSW, WRC, QLZ, QL, TR, and FL contributed to the interpretation of the data and critically revised the manuscript. QL, TR and FL contributed equally to the work as senior investigators. They designed the study, supervised the data analyses, and supervised the drafting and revising of the manuscript. The corresponding author attests that all the listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding statement
This study was supported by grants from the National Natural Science Foundation of China (82125032, 81930095, 81761128035, 82204048, and 82272079), the National Key Research and Development Program of China (2023YFE0109700), the Science and Technology Commission of Shanghai Municipality (23Y21900500, 19410713500, and 2018SHZDZX01), the Shanghai Municipal Commission of Health and Family Planning (GWV-11.1-34, 2020CXJQ01, and 2018YJRC03), the Shanghai Clinical Key Subject Construction Project (shslczdzk02902), Innovative research team of high-level local universities in Shanghai (SHSMU-ZDCX20211100), the Program of Shanghai Academic Research Leader (23XD1423400), the Shanghai Municipal Science and Technology Major Project (2018SHZDZX01 and 2021SHZDZX0103), and the Guangdong Key Project (2018B030335001).
Competing interests
The authors declare none.
Ethical standard
In the ABCD study, all procedures were approved by a central Institutional Review Board (IRB) at the University of California, San Diego, and in some cases by individual site IRBs (e.g. Washington University in St. Louis) (https://www.sciencedirect.com/science/article/pii/S1878929317300622). Parents or guardians provided written informed consent after the procedures had been fully explained and children assented before participation in the study.
Open access
