Many children with autism spectrum disorder (ASD) frequently display disruptive behaviors, including anger, disobedience, and refusal to comply with rules: prevalence estimates vary across behaviors and studies, with estimates of approximately 50% for temper tantrums, 30% for defiant behavior, and 20% for aggressive behavior (Chandler et al.,
2016; Kantzer et al.,
2018; Lecavalier,
2006; Maskey et al.,
2013; Simonoff et al.,
2008). Disruptive behaviors have repeatedly been related to unfavorable outcomes in children with ASD, e.g., social skills, to parental stress, and to decreased family functioning (Huang et al.,
2014; Matson et al.,
2010; Sikora et al.,
2013; Tomanik et al.,
2004). A vast body of research points to the importance of early-life environmental and child factors for ASD (Baio et al.,
2018; Charman & Chakrabarti,
2016; Hallmayer et al.,
2011; Mandy & Lai,
2016; Vijayakumar & Judy,
2016) and disruptive behaviors in children without ASD (Carneiro et al.,
2016; Guney et al.,
2015; Latimer et al.,
2012). However, studies on early-life and child factors related to disruptive behaviors in children with ASD are scarce.
The few available studies on early-life environmental factors associated with the presence of disruptive behaviors in ASD indicated involvement of pregnancy complications (Gadow et al.,
2008a) and parental variables such as mental health problems, parental stress, and less warm and more negative/harsh parenting (Bauminger et al.,
2010; Gadow et al.,
2008a; Maljaars et al.,
2014; McRae et al.,
2019; Midouhas et al.,
2013; Visser et al.,
2013). Furthermore, family demographic factors associated with disruptive behaviors in offspring with ASD may include single parenthood, low social economic status, large family size, and adolescent pregnancy (Chandler et al.,
2016; Gadow et al.,
2008a; Midouhas et al.,
2013; Simonoff et al.,
2008; St. Pourcain,
2011). Finally, child factors that have been related to disruptive behaviors in ASD include white ethnicity (Gadow et al.,
2008a) and intelligence (although inconsistent in its direction across studies; Cervantes et al.,
2014; Chandler et al.,
2016; Presmanes Hill et al.,
2014). However, existing research on disruptive behaviors in ASD often used small samples, was based on retrospective measurements, or investigated a limited set of factors.
In studies on early-life factors associated with the presence of ASD, disruptive behaviors are not consistently reported on or controlled for. Thus, some of the factors previously found associated with ASD may specifically be related to ASD with disruptive behaviors, and others to ASD without disruptive behaviors. Furthermore, it may be that factors associated with the presence of disruptive behaviors in children with ASD are also associated with the presence of ASD without disruptive behaviors. Currently, however, it is not well known how factors related to ASD with and without disruptive behaviors overlap.
In the current study, we investigated a broad range of early-life environmental and child factors (i.e., from the prenatal period until age 4) in relation to the presence of disruptive behaviors in children with autistic traits at age 7. Specifically, we looked at family demographic, pregnancy, delivery, neonatal, child, and preschool (0–4 years) parental and family factors, in a large sample of children from the prospective Avon Longitudinal Study of Parents and Children (ALSPAC; Boyd et al.,
2013; Fraser et al.,
2013). In order to investigate factors associated with the presence of disruptive behaviors in children with autistic traits, we compared children with autistic traits with and without co-occurring disruptive behaviors. We also made comparisons with unaffected controls, i.e., children without autistic traits and no disruptive behaviors, to investigate the overlap in factors associated with autistic traits
with and
without disruptive behaviors. We selected the variables in our model based on factors previously associated with disruptive behaviors in both ASD and non-ASD populations and factors previously associated with ASD, aiming to provide a synthesis of potential early-life factors associated with the presence of disruptive behaviors in ASD.
Results
Group Characteristics and Selection Bias
Table
2 shows the mean SCDC score (autistic traits) and SDQ
EXT score (disruptive behavior) around age 7, and, to describe the study sample, demographic characteristics per group in subset A (
n ASD + DB = 178,
n ASD − DB = 120,
n controls = 3,385) and B (
n ASD + DB = 307,
n ASD − DB = 198,
n controls = 5,896). As expected due to the use of cut-off scores, the mean SCDC score of participants in the unaffected controls group (i.e., children without autistic traits and no disruptive behaviors) was clearly below the mean score in both the ASD + DB group and the ASD – DB, and the mean SDQ
EXT score of the ASD + DB group was clearly above the mean of both the ASD − DB and the control group, in subset A as well as subset B. Additionally, the mean SCDC score was somewhat higher in the ASD + DB group than in the ASD − DB group, and the mean SDQ
EXT score was somewhat higher in the ASD − DB group than in the control group.
Table 2
Group characteristics in subsets A (participants with few missing values) and B (all participants), before imputation
SCDC: mean (SD) |
Aa | 13.0 (3.69) | 178 | 11.3 (2.75) | 120 | 1.80 (2.07) | 3,385 |
Bb | 13.5 (4.10) | 307 | 11.3 (2.82) | 198 | 1.76 (2.07) | 5,896 |
SDQEXT: mean (SD) |
Ac | 11.8 (2.20) | 178 | 6.00 (1.85) | 120 | 3.83 (2.27) | 3,385 |
Bd | 11.8 (2.25) | 307 | 6.06 (1.93) | 198 | 3.86 (2.28) | 5,896 |
Sex: % male |
A | 71.9 | 178 | 58.3 | 120 | 48.0 | 3,385 |
B | 73.0 | 307 | 59.6 | 198 | 48.5 | 5,896 |
IQ: mean (SD) |
A | 102 (17.8) | 137 | 104 (17.3) | 99 | 108 (15.5) | 2,817 |
B | 100 (17.2) | 206 | 103 (17.0) | 146 | 107 (15.3) | 4,519 |
Social class mothers: % high |
A | 39.4 | 155 | 41.2 | 102 | 43.7 | 2,997 |
B | 33.9 | 245 | 38.1 | 160 | 41.1 | 4,793 |
Social class partners: % high |
A | 38.6 | 158 | 50.0 | 102 | 47.5 | 3,117 |
B | 37.2 | 253 | 50.6 | 160 | 44.8 | 5,062 |
Family: % two-parent household |
A | 97.7 | 177 | 100 | 119 | 98.8 | 3,366 |
B | 93.5 | 292 | 96.3 | 189 | 95.5 | 5,673 |
Ethnicity: % white |
A | 96.6 | 178 | 98.3 | 119 | 97.5 | 3,364 |
B | 96.6 | 295 | 97.4 | 194 | 96.5 | 5,710 |
With regard to selection bias, the mean SDQEXT and SCDC score in the three groups did not differ between participants who were included in subset A (n = 3,683) and those who were excluded from subset A due to a high amount of missing values on independent variables (n = 2,718; n ASD + DB = 129, n ASD − DB = 78, n controls = 2,511), with the exception of the mean SCDC score in the ASD + DB group (included 13.0, excluded 14.2; t = 2.53, p = 0.01; see Online Resource Table 2). Participants included in subset A did not differ on sex distribution compared to participants who were excluded from this subset. Yet, included participants scored higher on IQ, were more often of white ethnicity, from a higher social class, and from a two-parent household (see Online Resource Table 3). This was also the case when comparing our total study sample (n = 6,401, i.e., subset B) with ALSPAC subjects that were excluded due to missing data on the SCDC and/or SDQ (8,031 subjects; see Online Resource Table 4).
Lasso Analyses
Component residual plots indicated that the assumption of linearity of the relationship between the independent variables and outcomes was not violated. No outliers were found (all Cooks Distance > 1) in any of the multiple imputations sets. Tables
3 and
4 show the results of the analyses: the independent variables which were found to be associated with higher (3) and lower (4) odds, around age 7, of ASD + DB compared to ASD – DB, ASD + DB compared to controls, and ASD − DB compared to controls, with
n (subset A/subset B) ASD + DB = 178/307,
n ASD − DB = 120/198, and
n controls = 3,385/5,896. A variable was assumed to be associated if it had a non-zero
b-coefficient in the lasso analysis in at least six out of ten multiple imputation sets in subset A and subset B (if included in B). To give an estimate of the magnitude of associations, the range (min–max) of odds ratios (
eb) across the multiple imputation sets of subset A and B together are displayed per variable. Odds ratios should be interpreted with caution since lasso creates an upward bias for non-zero coefficients, and thus also for the converted odds ratios. See Online Resource Table 1 for the scale range of the continuous variables. Online Resource Table 5 displays the range and number of non-zero
b-coefficients (i.e., change in log odds) per variable across the ten multiple imputation sets of subsets A and B, for each group comparison, and the range of the penalty terms (lambda) used in the lasso analyses.
Table 3
Lasso analyses: independent variables associated with higher odds of ASD + DB versus ASD − DB, ASD + DB versus controls, and ASD − DB versus controls, with the range of odds ratios across the multiple imputation sets of subset A (all variables; participants with few missing values) and B (variables with few missing values; all participants)
Family demographics | Partner’s medium social class (reference = high) | 1.001–1.411 | Maternal low social class (reference = high) | 1.043–1.704 | | |
Partner’s vocational education level (reference = CSE/none) | 1.007–1.525 | | | | |
| | Number of child’s siblings 18 months | 1.002–1.160 | | |
Pregnancy variables | Smoking mother | 1.013–1.370 | Smoking mother | 1.103–1.240 | | |
Smoking partner | 1.007–1.309 | Smoking partner | 1.029–1.229 | | |
Maternal antidepressant use | 1.171–3.241 | | | | |
| | Infections in second trimester | 1.019–1.115 | | |
| | Prenatal stress (life events) | 1.009–1.024 | | |
| | Premature birth | 1.001–1.237 | | |
| | High level of street traffic | 1.040–1.219 | | |
Delivery variables | Breech presentation at delivery/cesarean section* | 1.116–2.846 | | | | |
Neonatal variables | Feeding problems* | 1.054–2.257 | | | | |
Child variables | Male sex | 1.290–1.562 | Male sex | 2.199–2.532 | Male sex | 1.121–1.368 |
Temperament: adaptability (difficulty with waiting or change) | 1.015–1.067 | Temperament: adaptability (difficulty with waiting or change) | 1.043–1.075 | | |
Temperament: intensity (reacts strongly) | 1.038–1.061 | Temperament: intensity (reacts strongly) | 1.049–1.064 | | |
Temperament: persistence (reduced attention span) | 1.005–1.026 | Temperament: persistence (reduced attention span) | 1.024–1.038 | | |
| | Temperament: activity | 1.026–1.054 | | |
| | | | Temperament: mood (negative mood) | 1.005–1.040 |
Preschool parental and family variables | Negativity scale score 47 months mother | 1.034–1.068 | Negativity scale score 47 months mother | 1.230–1.273 | Negativity scale score 47 months mother | 1.147–1.192 |
Negativity scale score 47 months partner* | 1.001–1.073 | Negativity scale score 47 months partner* | 1.137–1.203 | Negativity scale score 47 months partner* | 1.031–1.095 |
| | | | Positivity scale score 47 months partner* | 1.046–1.214 |
| | Parental bonding 8 months partner | 1.007–1.039 | | |
Maternal antisocial behavior | 1.001–1.099 | Maternal antisocial behavior | 1.091–1.156 | | |
| | Child maltreatment | 1.171–1.297 | Child maltreatment | 1.065–1.206 |
| | Aggression between parents 8 months (mother-rated) | 1.002–1.052 | Rows between parents 33 months (partner-rated)* | 1.028–1.057 |
| | Maternal depression | 1.003–1.023 | Maternal depression | 1.013–1.040 |
| | | | Partner’s anxiety | 1.010–1.050 |
| | | | Stress (life events) | 1.002–1.013 |
| | | | Accident prevention measures (in household) | 1.013–1.074 |
Table 4
Lasso analyses: independent variables associated with lower odds of ASD + DB versus ASD − DB, ASD + DB versus controls, and ASD − DB versus controls, with the range of odds ratios across the multiple imputation sets of subset A (all variables in participants with few missing values) and B (variables with few missing values in all participants)
Family demographics | Summer birth | 0.715–0.836 | | | | |
| | Twin birth | 0.474–0.743 | | |
Pregnancy variables | Weekly seafood consumption | 0.733–0.971 | Weekly seafood consumption | 0.662–0.885 | | |
Diabetes | 0.499–0.908 | | | | |
| | Vaginal bleeding in pregnancy | 0.772–0.957 | | |
Delivery variables | | | | | Breech presentation before labor* | 0.816–0.997 |
Neonatal variables | Anemia | 0.750–0.950 | Anemia | 0.754–0.907 | | |
| | | | Feeding problems* | 0.760–0.903 |
Child variables | Temperament: approach (difficulty with the unknown) | 0.981–0.999 | Temperament: approach (difficulty with the unknown) | 0.989–0.997 | | |
Temperament: threshold (quicker to notice things) | 0.948–0.967 | Temperament: threshold (quicker to notice things) | 0.944–0.964 | | |
| | Intelligence | 0.978–0.993 | Intelligence | 0.986–0.997 |
Preschool parental and family variables | Positivity scale score 47 months mother | 0.880–0.929 | Positivity scale score 47 months mother | 0.815–0.891 | | |
Positivity scale score 47 months partner* | 0.762–0.895 | Positivity scale score 47 months partner* | 0.891–0.994 | | |
Parenting score 6 months mother | 0.833–0.882 | Parenting score 18 months partner | 0.983–0.997 | | |
Parental bonding 33 months mother | 0.966–0.987 | Parental bonding 33 months mother | 0.924–0.957 | Parental bonding 33 months mother | 0.973–0.991 |
| | Warmth between parents 33 months (mother-rated) | 0.972–0.992 | Warmth between parents 33 months (mother-rated) | 0.989–0.999 |
| | | | Affection between parents 8 months (partner-rated)* | 0.984–0.999 |
Accident prevention measures (in household) | 0.914–0.995 | | | | |
Explained Deviance
The results of the ASD + DB versus ASD – DB model showed that explained deviance, i.e., the total fit of the model, ranged from 22.5 to 30.6% in subset A (
n = 298) and from 16.0 to 19.6% in subset B (
n = 505) across the ten imputation sets. In the ASD + DB versus controls model, independent variables explained 33.1 to 35.1% of deviance in subset A (
n = 3,385) and 29.2 to 31.2% in subset B (
n = 6,203). The ASD – DB versus controls model explained 7.6 to 12.4% of deviance in subset A (
n = 3,505) and 6.6 to 8.7% in subset B (
n = 6,094). Table
5 shows the percentage of explained deviance per variable category in each group comparison in subset A. Of note, these percentages do not add up to the explained deviance in the total model due to interrelationship of variables (i.e., leaving out one variable category can lead to other variable categories taking over explained deviance) and should thus be seen as estimations.
Table 5
Mean explained deviance in percentages per variable category in the models comparing ASD + DB versus ASD − DB, ASD + DB versus controls, and ASD − DB versus controls, across the ten imputation sets in subset A
Family demographics | 1.39 | 0.81 | 0.12 |
Pregnancy | 3.59 | 1.13 | 0.49 |
Delivery | 0.76 | 0.11 | 0.65 |
Neonatal | 0.68 | 0.11 | 0.17 |
Child | 5.60 | 7.45 | 1.67 |
Preschool parental and family | 6.32 | 11.66 | 5.33 |
Family Demographics
Medium social class and vocational education level during pregnancy, both of the mother’s partner, were associated with higher odds of ASD + DB around age 7 in comparison to ASD – DB. Summer birth was associated with lower odds of being in the ASD + DB group versus the ASD − DB group. A lower maternal social class during pregnancy and a higher number of child’s siblings at 18 months were found to be associated with higher odds of ASD + DB versus controls, while twin birth was associated with lower odds of ASD + DB versus controls.
Pregnancy, Delivery, and Neonatal Variables
With regard to pregnancy variables, smoking during pregnancy (of the child’s mother and her partner) was associated with higher odds of ASD + DB around age 7 versus ASD − DB and versus controls. Eating seafood on a weekly basis was associated with lower odds of ASD + DB versus ASD – DB and versus controls. Maternal antidepressant use was associated with higher odds of ASD + DB compared to ASD − DB. In contrast, maternal diabetes was associated with lower odds of ASD + DB versus ASD − DB. Maternal infections, prenatal stress (life events), a high level of street traffic, and being born prematurely were associated with higher odds of ASD + DB versus controls, whereas vaginal bleeding was associated with lower odds of ASD + DB versus controls.
Of the measured delivery and neonatal variables, a breech presentation and feeding problems in the first two weeks after birth were associated with higher odds of ASD + DB versus ASD – DB, whereas breech presentation and feeding problems were associated with lower odds of ASD − DB versus controls. Neonatal anemia was associated with lower odds of ASD + DB versus ASD − DB and versus controls.
Child Variables
Male sex was associated with higher odds of ASD + DB versus ASD − DB around age 7, and versus controls, but also with higher odds of ASD − DB versus controls. Of preschool temperament characteristics, measured at 24 months, a temperament characterized by less difficulty with the unknown, being less quick to notice things, more difficulty with waiting or change, stronger reactions, and a reduced attention span were all associated with higher odds of ASD + DB versus ASD − DB and versus controls around age 7. In addition, a temperament characterized by more activity was related to higher odds of ASD + DB versus controls. In contrast, a temperament characterized by a more negative mood was associated with higher odds of ASD − DB versus controls. Higher intelligence was associated with lower odds of ASD + DB versus controls and with lower odds of ASD − DB versus controls.
Preschool Parental and Family Variables
Of the preschool parental and family variables, a more negative parent–child relationship (i.e., negativity scale of mother as well as her partner) and maternal antisocial behavior were associated with higher odds of ASD + DB around age 7 versus ASD − DB and versus controls. A more positive parent–child relationship (i.e., positivity scale), a higher degree of maternal enjoyment and confidence (bonding), and a higher frequency of parent–child activities (parenting score) were associated with lower odds of ASD + DB versus ASD − DB and versus controls. A more negative parent–child relationship was also associated with higher odds of ASD − DB versus controls, while maternal bonding was also associated with lower odds of ASD − DB versus controls. The use of more accident prevention measures in the household was associated with lower odds of ASD + DB versus ASD − DB and with higher odds of ASD – DB versus controls.
A worse psychological health of parents in the child’s preschool years, a less optimal relationship between parents (i.e., more aggression/less affection), and more child maltreatment were associated with higher odds of ASD + DB around age 7 versus controls and with higher odds of ASD − DB versus controls. Stress during the child’s preschool years (life events) was associated with higher odds of ASD − DB versus controls. In comparison to controls, there were also positive parent–child factors associated with higher odds of ASD + DB and ASD − DB, i.e., mother’s partner’s bonding to the child was associated with higher odds of ASD + DB and a more positive relationship between partner and child was associated with higher odds of ASD − DB.
Discussion
We examined the association of a wide range of early-life factors up to age 4 with the presence of disruptive behaviors in children with autistic traits around age 7. Investigation of factors that discriminate between children with autistic traits with versus without disruptive behaviors, indicated that maternal antidepressant use, prenatal smoking, not consuming seafood during pregnancy, a low prenatal social economic situation, a breech presentation at delivery, neonatal feeding problems, child male sex, child difficult temperament, a less optimal preschool family environment, and maternal antisocial behavior were all associated with the presence of disruptive behaviors in children with autistic traits.
We found that many of above early-life factors were only associated with the presence of disruptive behaviors in children with autistic traits, but not with the presence of autistic traits without disruptive behaviors in comparison to unaffected controls (i.e., children without autistic traits and no disruptive behaviors). This was also apparent from the estimated total level of fit of the tested models, that lay around 20% for the ASD + DB versus ASD − DB model, but around 10% for the ASD – DB versus controls model.
Prenatal parental smoking was one variable that was associated with the presence of disruptive behaviors in children with autistic traits, but not with the presence of autistic traits without disruptive behaviors, a finding that is in line with other studies that did not find an association between smoking and ASD (Rosen et al.,
2015; Wang et al.,
2017). A study that did point to a relation between prenatal smoking and ASD (St. Pourcain et al.,
2011) included children with ASD, who mainly also had externalizing problems. It may thus be that the association in that study stemmed from an association between smoking and ASD with co-occurring disruptive behaviors. Also regarding child temperament, measured at 24 months, we found many aspects (e.g., impaired adaptability, intensity, persistency, and approach) to be associated with the presence of disruptive behaviors in children with autistic traits around age 7, while only one temperamental aspect (i.e., a more negative mood) was associated with the presence of autistic traits without disruptive behaviors in comparison to unaffected controls. Our findings are in in line with studies that suggest that early temperament may be a vulnerability for later disruptive behavior (Nigg,
2006; Rijlaarsdam et al.,
2016).
Just as prenatal smoking and a difficult child temperament, not consuming seafood during pregnancy, a breech presentation at delivery, neonatal feeding problems, and a low prenatal social economic situation were related to the presence of disruptive behaviors in seven-year-old children with autistic traits, but not to the presence of autistic traits without disruptive behaviors. In addition, in comparison to unaffected controls, maternal infections during pregnancy, traffic related air pollutants, premature birth, and a large family size were associated with autistic traits with disruptive behaviors, but not with autistic traits without disruptive behaviors. This finding thus also indicates a specific association of these factors with ASD
with disruptive behaviors and not
without, even though no effect was found in comparison of the two ASD groups themselves. Our findings on low social economic situation, large family size, and pregnancy related complications correspond to studies in which these factors were associated with disruptive behaviors in ASD (Chandler et al.,
2016; Gadow et al.,
2008a; Midouhas et al.,
2013). In other studies, however, above factors have been associated with ASD (Bölte et al.,
2019; Emberti Gialloreti et al.,
2019; Gardener et al.,
2011; Julvez et al.,
2016; Wang et al.,
2017). Given our findings, it may be that findings in previous studies on ASD were driven by co-occurring disruptive behaviors.
In contrast to the factors described above, other factors were associated with the presence of disruptive behaviors in children with autistic traits, but also with the presence of autistic traits without disruptive behaviors in comparison to unaffected controls: child male sex, maternal depression, and a less optimal preschool family environment. Yet, in previous studies (Brereton et al.,
2006; Chandler et al.,
2016; Simonoff et al.,
2008), male sex was not found to be related to disruptive behaviors in ASD. The high number of girls in our ASD groups (about 33% versus less than 20% in other studies) may have provided more room for finding sex differences. This high number of girls may stem from defining the ASD groups based on high levels of autistic traits, instead of a clinical diagnosis. Since girls with ASD are less likely to receive a diagnosis than boys (Lai & Baron-Cohen,
2015), our results emphasize the value of studying ASD as the extreme of an autistic trait dimension (Constantino & Charman,
2016; Robinson et al.,
2011; Ronald & Hoekstra,
2011). The association that we found between male sex and autistic traits without disruptive behaviors is in line with previous studies into ASD (Baio et al.,
2018; St. Pourcain et al.,
2011).
Regarding maternal depression, we found that maternal prenatal antidepressant use was related to the presence of disruptive behaviors in children with autistic traits and that maternal depressive symptoms were associated with the presence of autistic traits without disruptive behaviors in comparison to controls. Since the role of antidepressants can be (partially) explained through maternal mental illness (Andrade,
2017; Rai et al.,
2017; Rotem-Kohavi & Oberlander,
2017), our results suggest that maternal depression may be related to ASD and, above that (shown through the association of antidepressants), to disruptive behaviors in children with ASD. This is in line with other studies showing parental psychopathology to be related to disruptive behaviors in ASD (Gadow et al.,
2008a; McRae et al.,
2019) and with studies indicating no direct effect of antidepressants on externalizing behavior (see Rotem-Kohavi & Oberlander,
2017).
Last, a suboptimal preschool family environment was also associated with the presence of disruptive behaviors in children with autistic traits around age 7 as well as with the presence of autistic traits without disruptive behaviors compared to controls. Of preschool family environment factors, fewer parent–child activities and fewer household accident prevention measures were specifically associated with the presence of disruptive behaviors in children with autistic traits, while a worse parent–child relationship, and less parental enjoyment and confidence in parenting were associated with disruptive behaviors in children with autistic traits, but also with autistic traits without disruptive behaviors. These findings are in line with other studies that suggest an association between a less optimal early parent–child relationship and disruptive behaviors in children with ASD (Maljaars et al.,
2014; McRae et al.,
2019; Midouhas et al.,
2013), as well as with ASD (Mandy & Lai,
2016). Our results indicate that a worse family environment during the preschool period is associated with the presence of autistic traits
with and
without disruptive behaviors at child’s later age, but that this association may apply more strongly to autistic traits
with disruptive behaviors.
In this study, we did not look into factors associated with the presence of disruptive behaviors in children
without autistic traits. This was because our focus was on children with autistic traits: adding a fourth ‘DB-only’ group would make the manuscript lose its focus. Furthermore, already quite some studies have been done on factors associated with disruptive behaviors in non-ASD children. When comparing our findings on factors associated with the presence of disruptive behaviors in children with autistic traits to the literature on factors associated with disruptive behaviors in non-ASD populations, it is interesting to see that many factors coincide. This is consistent with previous studies that suggest similarities in factors associated with disruptive behaviors in ASD and non-ASD populations (Gadow et al.,
2008b; Mandy et al.,
2014). Coinciding factors include prenatal smoking (Dolan et al.,
2016; Latimer et al.,
2012; Ruisch, Buitelaar, et al.,
2018; Ruisch, Dietrich, et al.,
2018), low SES (Carneiro et al.,
2016; Ruisch, Buitelaar, et al.,
2018), maternal antisocial behavior (Davies et al.,
2012), maternal depression (Guney et al.,
2015), a less optimal preschool family environment (Gach et al.,
2018), child male sex (Carneiro et al.,
2016), and child difficult temperament (Bao et al.,
2016; Carneiro et al.,
2016).
Finally, our results showed that a few child and environmental factors were related to both ASD groups (with and without disruptive behaviors) in comparison to unaffected controls. These were lower child intelligence, preschool child maltreatment, and a bad relationship between parents. Furthermore, stress, which was measured by number and impact of life events, was also associated with both ASD groups, albeit at a different age: stress during pregnancy was associated with the presence of autistic traits with disruptive behaviors around age 7, while environmental stress during the preschool period was associated with the presence of autistic traits without disruptive behaviors. The period in which stress occurs may thus be important. In another study, maternal stress during pregnancy was not associated with ASD, in contrast to stress during preschool years (Rai et al.,
2012). In line with our results on prenatal stress, maternal stress during pregnancy has also been related to externalizing behavior in non-ASD populations (Latimer et al.,
2012; MacKinnon et al.,
2018; Ruisch, Buitelaar, et al.,
2018).
Overall, our results show that some early-life environmental and child factors in our study are associated with autistic traits regardless of the presence of disruptive behaviors around age 7, but that more of the studied early-life and child factors are related to autistic traits with disruptive behaviors. The factors that were related to the presence of disruptive behaviors in children with autistic traits came from all investigated variable categories: family demographics, pregnancy, delivery, neonatal, child, and preschool parental and family factors. Our results suggest that pregnancy, child, and preschool parental and family factors may be especially related to the presence of disruptive behaviors in children with autistic traits around age 7. With regard to the clinical implications of our study, our results indicate that it is important for clinicians and parents to be aware that early-life environmental and child factors may be associated with the presence of disruptive behaviors in a child with autistic traits or ASD at a later age.
Strengths and Limitations
Strengths of our study include the prospective measurement of a comprehensive range of factors (thus eliminating recall bias) in a large population sample. The longitudinal design of ALSPAC provided us with the opportunity to examine the association of early-life environmental factors with autistic traits and disruptive behaviors at a later age, i.e., around age 7. However, we cannot draw conclusions about the associations of these factors with disruptive behaviors and autistic traits measured closer to the age at which the factors were measured. Future research may look into these concurrent associations as well as the possible reciprocity of relationships of the studied factors with disruptive behaviors and autistic traits.
Defining our ASD groups by means of an autistic traits cut-off, instead of a clinical diagnosis, may be seen as a limitation. However, our SCDC cut-off has adequate discriminative value for ASD (Skuse et al.,
2005), and we therefore think that our findings may be extended to children with a clinical ASD diagnosis. Furthermore, using a traits cut-off may also have advantages, leading to inclusion of children who are less likely to receive a clinical diagnosis, for example girls (Constantino & Charman,
2016). In future research it would be interesting to see how the variables found in relation to autistic traits hold when studied in children with a clinical diagnosis of ASD.
Another limitation is that, even with many variables included, our study may have overlooked variables of interest for ASD or disruptive behavior, e.g., prenatal maternal valproate use, nutrition, substance use, and parenting style. Furthermore, our study design (i.e., lasso analyses) was chosen to indicate whether a specific variable was or was not associated, and thus not to indicate the strength of individual associations. Future research may look further into the strength of the associations found in the current study.
Finally, the ASD + DB group scored higher on autistic traits than the ASD − DB group and the ASD − DB group scored higher on disruptive behaviors than the control group. The associations with regard to the presence of disruptive behaviors in children with autistic traits may thus partly stem from higher levels of autistic traits in the ASD + DB group, although, for most associated factors this does not seem likely since these factors were not associated with the presence of autistic traits without disruptive behaviors. Future research could aim to more thoroughly disentangle the effects of more severe autistic traits from co-occurring disruptive behaviors.