Background
Autism spectrum disorder (ASD) is a condition affecting 1–2% children worldwide [
1]. Individuals with ASD are impaired in reciprocal social communication and interaction and display various stereotyped and repetitive behaviors [
2]. Many children with ASD also meet the criteria for attention-deficit/hyperactivity disorder (ADHD), with rates of approximately 30–60% in community samples [
3‐
5] compared to 5–7% for ADHD in the general population [
6]. Both ASD and ADHD are associated with highly overlapping cognitive impairments (e.g., [
7‐
9]). Therefore, neurocognitive approaches can be useful for explaining the mechanisms underpinning the co-occurring ADHD in the ASD population [
10‐
12].
Two disorders can co-occur beyond chance level due to artifactual or non-artifactual reasons [
13,
14]. In the clinical setting, the co-occurrence of two disorders can arise due to sampling or ascertainment biases, resulting in artifactual increase of its prevalence. However, these mechanisms cannot be the sole explanation for the co-occurrence of ASD and ADHD because their high rates have been observed in population-based and epidemiological samples [
3‐
5]. Indeed, several models of non-artifactual comorbidity of ASD and ADHD have been proposed recently [
10‐
12,
15]. The “additivity model” is one such model [
11]. The model hypothesizes that the co-occurrence of ASD and ADHD arises from separate but correlated risk factors or liabilities and results in the “additive combination of two separate nosologies” [
14]. To investigate this model, neurocognitive studies typically use group comparison or factorial design, contrasting individuals with pure ASD or ADHD, combined ASD + ADHD, against those with typical development. In these studies, the pure ASD or ADHD groups are expected to demonstrate unique cognitive profiles, of which combination characterizes the cognitive performance of the ASD + ADHD group. The model thus predicts a double dissociation between the cognitive correlates of ASD and ADHD traits [
14].
In the context of cognitive function, impairments of theory of mind (ToM) and executive function (EF) cognitive domains are often reported in children with ASD (e.g., [
9,
16‐
19]). Central to ToM is the ability to mentalize, that is, to attribute mental states such as beliefs, desires, feelings, and intentions of others. Impaired ToM abilities are thought to be developmentally specific to ASD [
20,
21] and have been reported to much lesser extent in ADHD (e.g., [
9,
22,
23]). Often reported in both disorders are EF impairments, which are usually inferred from performance of inhibition, working memory, cognitive flexibility, and planning tasks (see e.g., [
24,
25]). The impairments of EF have been reported especially in ADHD [
26‐
28], and it is uncertain presently if executive dysfunctions truly characterize ASD. Findings of impairments across studies have been heterogeneous (e.g., [
17,
29,
30]), and the performance of the EF tasks are rarely correlated with ASD symptom severity, except for between cognitive flexibility deficits and repetitive behavior (e.g., [
31‐
34]). Evidence also suggests that EF impairments in the ASD population are associated with co-occurring ADHD traits. For instance, Corbett et al. [
35] found that the inhibitory impairments among ASD children fell to trend level after excluding those with additional ADHD, while Buehler et al. [
36] showed increasing motor inhibition impairments in the ASD + ADHD relative to the pure ASD group. Others have also reported increased impairments of sustained attention and working memory in the combined vs. the pure ASD group [
11,
12,
15,
37,
38].
Those findings suggest dissociable relations between EF and ToM impairments and the ADHD or ASD traits among individuals with ASD [
39], but there are limitations to this interpretation. Firstly, the pattern of increased executive dysfunctions in the combined relative to the pure ASD group have not been observed in
every EF subdomain (e.g., in the domains of planning and cognitive flexibility [
11,
40,
41]), although presumably, these subdomains have specific characteristics beyond the “common” EF [
25,
42] that are unrelated to the ADHD traits. Furthermore, most studies tend to use single measures for defining symptoms of ASD and ADHD, completed typically by a single informant (e.g., parents for pediatric studies). Such approach may increase the magnitude of associations between symptom domains because of the shared source of information, in this case between, ASD and ADHD symptoms [
39]. Finally, most studies have tested the links between each domain of cognition (ToM or EF) and behavioral traits in separate studies [
39]. Thus, possible influences each factor might have on the relation between the other cognitive domain and behavioral traits might be obscured. Moderating influences could exert for instance through pathways between EF and ToM (e.g., [
43‐
46]) or between the symptom domains of ASD and ADHD (e.g., [
47‐
49]), which should be controlled within one single framework.
We explore the associations among EF, ToM, and symptom domains of ASD and ADHD and address the above challenges in the present study. In line with a recent approach [
39], we investigated the dimensional relations between factors within a population-based cohort of children with ASD instead of conducting a group comparison with a factorial design. The study used a structural equation modeling (SEM) framework, enabling the constructions of the latent factors EF and ToM derived from multiple measures, and therefore controlling the specific influences of each task or cognitive subdomain. Symptoms ASD and ADHD were indexed using multiple measures collected from multiple informants including parents, teachers, and the children themselves. Using this approach, we aimed to elucidate the specific and potentially clinically significant patterns of associations among those factors within the ASD population. We hypothesized specific associations between EF impairments and increased ADHD symptoms, and between ToM impairments and increased ASD symptoms.
Discussion
Although ADHD symptoms are frequently seen in people with ASD, the basis for this association remains poorly understood. This study explored the specificity of relations among executive dysfunction, ToM impairments, and the ASD and ADHD traits using the framework of SEM, in a population-based sample of children with ASD. The primary findings of the study are that poorer EF ability is specifically associated with increased ADHD symptoms, accounting for its association with variation in ToM performance and ASD symptoms. Furthermore, impairments in ToM are specifically associated with ASD symptoms. Taken together, the findings show that ADHD and ASD symptoms in adolescents with ASD have dissociable neurocognitive correlates. The secondary findings are that the observed correlation between ASD and ADHD symptoms in the sample can be explained by shared source of reporting from parents and the correlation between EF and ToM. Finally, the specific relations between each cognitive domain and the behavioral symptoms are retained even when controlling for IQ.
Specific associations between cognitive domains and symptoms
Our model shows that the EF impairments among individuals with ASD are associated with increased ADHD symptoms. This is in line with findings from previous group comparisons reporting increased EF difficulties among children with dual diagnoses of ASD and ADHD compared to children with ASD alone [
11,
36‐
38] and the associations reported between EF deficits and increased ADHD symptoms in ASD samples [
39,
83]. In addition, the association between EF impairments and increased ADHD symptoms in this ASD sample is consistent with findings in other samples of individuals with ADHD (e.g., [
8,
26,
27,
84]) and in the general populations (e.g., [
85‐
87]). Importantly, we did not detect a significant association between ToM impairments and ADHD symptoms in the model, consistent with a recent meta-analytic finding showing that ToM difficulties are present to a much lesser extent in ADHD than ASD populations [
9]. The specific relation between EF and ADHD symptoms in this context may provide an explanation for the mixed findings of EF difficulties in ASD (e.g., [
17,
29,
30]), in that, EF difficulties are perhaps more likely to be found among people with ASD who have co-occurring ADHD.
The SEM model also shows that mentalizing abilities are specifically associated with the severity of ASD and not ADHD symptoms. This is consistent with results from previous studies [
88‐
90] and supports the view that ToM impairments are specifically linked to ASD symptoms. These findings contrast with some studies that fail to find an association between ToM and everyday social behavior in people with ASD (e.g., [
91‐
93]), perhaps due to these studies’ reliance on specific measures such as the false belief test, which may not fully capture the breadth of socio-cognitive and perceptual processes related to ToM [
92,
94,
95]. We have addressed this potential limitation by using a multi-measure approach to better capture the ToM construct.
Contrary to previous findings of associations between performance of a variety of EF tasks and ASD symptoms [
39,
44,
92,
96] and specifically between cognitive flexibility and repetitive behavior symptoms (e.g., [
31‐
34]), no association was found between EF impairments and ASD traits in the SEM model. There could be several explanations for this finding. Firstly, none of these previous studies considered the association between EF and ASD after controlling for ToM impairments. Therefore, the association between EF and ASD symptoms might have been observed in those studies because the covariation of ToM and ASD or EF and ADHD symptoms was not accounted for. Secondly, many studies reporting the associations between EF impairments and ASD symptoms analyzed these relations separately between tasks that may tap different aspects of EF. Thus, the correlations found between task performance and ASD symptoms, notably between cognitive flexibility performance and repetitive behavior symptoms, may reflect relations between properties that are specific to the task or the EF subdomain, rather than the underlying common EF factor, with the ASD symptoms. Lastly, as learnt from the finding in relation to the Luria Hand Game, performance on a task that is traditionally an inhibitory measure may load onto ToM instead of EF factors, thus more associated with ASD symptoms. This could be because the task triggers mentalizing processes when the subjects attempt to guess which hand gesture the experimenter would be giving next. Such interpretation is in line with the Triple-I hypothesis [
97], which argues that EF impairments in ASD might be a by-product of mentalizing deficits in the population.
A shared parental rating factor partially accounts for the correlation between ASD and ADHD traits
Several previous studies have found that individuals with ASD (and co-occurring ADHD symptoms) have more severe ASD traits relative to those with ASD alone, judged from parent-rated questionnaires such as the SRS [
38,
47,
49]. Autistic trait measures have less specificity when applied to children with ASD and additional behavioral or emotional problems, including ADHD [
98,
99]; thus, it is possible that past reports of an association between ASD and ADHD traits were due to a systematic instrument bias, in which those with additional ADHD traits also receive higher ratings of autistic traits. The association between ADHD and ASD trait severity was noticeably absent in previous studies; however, when ASD traits were measured using measures such as the ADI-R, ADOS [
38,
47,
48,
75] or clinical symptom counts [
74]. Unlike the SRS, which is parent questionnaire, measures such as the ADI-R and clinical symptoms involve clinical judgment, and in the case of ADOS, direct observation of the children. Therefore, the association between the ASD and ADHD traits found in previous studies might be partially dependent on the source and type of reporting.
Indeed, the final SEM model showed that information obtained from the same source (i.e., parents in this case) on ASD and ADHD traits moderated their correlation. That is, ratings by the same informant on different measures are more highly correlated than those from different informants. There are multiple possible explanations for this finding. Informants, in this case, parents, may have a specific response style, such as a tendency to rate all behaviors as high or low, that influences their responses across measures [
100]. Furthermore, children’s behavior may differ across settings [
101,
102]. As in many studies, parents were the predominant source of information regarding symptoms in this study. Because no other information source was shared across ASD and ADHD measures, it was not possible to test whether these effects are specific to parents or applies to other sources, such as teachers or direct observation. Nevertheless, our findings underline the importance of obtaining multiple sources of information [
103].
Explaining the co-occurrence of ADHD symptoms among ASD children
In line with the comorbidity literature, our model showed that artifactual and non-artifactual mechanisms could explain the increased co-occurrence of ADHD traits in the ASD children population. Firstly, ADHD traits might be reported at increased rates, presumably in those with more severe ASD, due to shared reporting source effects from parents. These cases, which do not express true comorbidity, one hopes are few and can be differentiated from true comorbidity cases during clinical observations, corroborated by non-parental source of information. Aside from this artifactual reason, our model provides preliminary supports for the cognitive mechanisms underlying the true comorbidity of ADHD traits in children with ASD. We have shown in this study that the cognitive correlates of the ASD and ADHD traits in children are dissociable, which supports the additive model for the co-occurrence of ADHD traits on ASD symptoms. Furthermore, from our model, the co-occurrence of the separable ASD and ADHD traits could be explained partially by the moderate correlation between EF and ToM impairments (.56), indicating that individuals with both ASD and ADHD symptoms constitute those who are “doubly hit” by both EF and ToM impairments.
Due to the limitations posed by the available data, the support our model gives to the additive model is preliminary. To satisfactorily adhere to the additive comorbidity model, we believe that the ADHD traits found in our ASD population must be phenomenologically equivalent, both in presentations and their associations to the cognitive factors, with symptoms found in the pure ADHD population. While previous findings suggest that ADHD symptoms in clinical ASD and ADHD populations have similar presentations (inattentive, hyperactive, and combined [
104,
105]) with only subtle differences in few symptoms [
106,
107], and although the association found between executive dysfunction and ADHD traits in our sample is consistent with other findings in ADHD population (e.g., [
8,
26,
27,
84]), neither could be directly observed in this study due to the lack of ADHD comparison sample.
Strengths and limitations
The inclusion of a well-characterized sample of children with ASD is an advantage for this study. Although the sample is not ASD population representative due to the exclusion of individuals with IQ below 50, it is still population-based and is free from clinical sampling biases. In fact, the comparisons we made between eligible individuals who did and did not take part in the study suggested that the study participants had representative ASD and ADHD traits for ASD children with IQ of 50 and above, therefore substantively extending previous neurocognitive findings in this topic which typically excluded children with learning disability. The inclusion of multiple measures from multiple informants gives an advantage over the typical approach of using single measures. The study includes multiple measures of ADHD psychopathology including the SDQ, a well-recognized screening instrument for children and adolescents, and the diagnostic instrument CAPA. Furthermore, the ADOS and the ADI-R, used to index autistic traits in the model, are considered “gold-standard” research instruments for assessing ASD [
108].
One limitation of the study is that factors were modeled with EF and ToM predicting the ADHD and ASD symptoms, in line with the idea that EF and ToM deficits are endophenotypes that mediate the link between genes and behavioral symptoms (see [
10] for alternative models that do not assume this directionality). However, the study design does not allow a strong test of causal direction. Therefore, the links between factors in the present model are best understood as associations rather than predictions. Secondly, we have modeled associations between factors indexed by measures collected over a 6-year period, thus assuming the stability of behavioral traits within the time window, which was supported by findings of persisting ASD and ADHD traits from childhood to adolescents reported in previous studies [
74‐
76]. Furthermore, the additional analyses including only data collected during the adolescent years preserve the specificity of relations between the factors EF and ADHD and between ToM and observed ASD symptoms in the model. Finally, the small number of girls in the study may limit the generalizability of the results among females with ASD.