This study is a secondary analysis of previously collected data from a social skills training group intervention for adolescents and youths with ASD [30]. In the original study, participants were screened out if they had 1) a verbal IQ score lower than 65 on the Wechsler Adult Intelligence Scale (WAIS-IV), Korean Educational Development Institute-Wechsler Intelligence Scale for Children-Revised (KEDI-WISC), or the Wechsler Intelligence Scale for Children (WISC-IV); 2) a history of major mental illness (e.g., schizophrenia or bipolar disorder), 3) severe oppositional and aggressive tendencies; or 4) hearing, visual, or physical disabilities or neurological illnesses that would prevent participation in the intervention program. Among 110 participants who attended the social skills intervention, 14 were excluded from this study because we did not have data to confirm their diagnosis.

A total of 96 participants with ASD [age range=11-18 yearsold; mean age (standard deviation)=14.30 years (1.80); 91 males] were included in this study. Detailed demographic characteristics are reported in Table 1. All data collection procedures (IRB no: B-1108/133/007, B-1301/187-006, B-1106/129-009) and retrospective analyses of collected data (IRB no: B-2001/591-103) were approved by the Institute Review Board (IRB).

Participants were recruited from the child and adolescent psychiatric clinics of the corresponding author’s institution and from advertisements mailed to child psychiatrists registered in the Korean Academy of Child and Adolescent Psychiatry or shared on online parent support group blogs. Once the participants and their parents had given their informed consent, the Korean translation of the Autism Diagnostic Observation Schedule-Second Edition (ADOS-2) [31,32] the Korean translation of the Autism Diagnostic Interview-Revised (ADI-R) [33,34], and the WAIS or WISC were administered during their one-time visit to the authors’ laboratory to screen for eligibility before the intervention. Research-reliable researchers (i.e., trained professionals who are qualified to implement the ADOS-2 and/or ADI-R) administered the ADOS-2 and ADI-R, and trained graduate students administered and scored the WAIS or WISC under the close supervision of clinical psychologists. All participants scored higher than the cutoff points in both the ADOS-2 and ADI-R diagnostic algorithms. In addition, all caregivers of participants completed a battery of surveys including the Social Communication Questionnaire (SCQ) [35] and the Social Responsiveness Scale-Second Edition (SRS-2) [36] to confirm the participants’ diagnoses. The ASD diagnoses of all participants were confirmed by psychiatrists’ best clinical judgments based on their ADOS-2, ADI-R, SCQ, and SRS-2 scores.

A few weeks later, eligible participants came back to participate in the intervention. After the first session of the intervention, participants and caregivers completed the State/Trait Anxiety Inventory (STAI) [37] and the Child Behavior Checklist for Ages 6-18 (CBCL/6-18) [38]. See Yoo et al. [30] for a detailed account of the recruitment and screening procedures and detailed information about the intervention.

First, the means, standard deviations of, and zero-order correlations among the STAIT, STAIS, and CBCL-Anxiety were computed. The equality of the correlation coefficients of the STAIT and CBCL-Anxiety and of the STAIS and CBCL-Anxiety was tested, using the “cortesti” function (i.e., a function that allows the approximation of whether a certain correlation strength is different from the strength of another correlation) in Stata statistical software: Release 15 (College Station, TX, USA) [49]. Then, pairwise Pearson r correlations among STAIS, STAIT, CBCL-Anxiety, age, and FSIQ were calculated. Subsequently, partial correlations among the STAIT, STAIS, CBCL-Anxiety, total RRB algorithm score from the ADI-R, and total scores from the SRS Mannerism subscales were computed, while controlling for age and FSIQ. To identify the specific types of RRBs associated with anxiety, we conducted partial correlations among STAIT, STAIS, CBCL-Anxiety, the four subdomain scores of the RRB domain, and the sensory sensitivity variable of ADI-R. Sensory sensitivity was calculated by adding the three ADI-R items that capture participants’ atypical responses to sensory stimuli. Finally, the partial correlations among Withdrawn/Depressed, Somatic Complaints, Rule-Breaking, Aggressive Behavior subscales of the CBCL, STAIT, STAIS, and CBCL-Anxiety were computed, while controlling for age and FSIQ.

First, age was associated with state anxiety but not with selfreported trait anxiety, showing that trait anxiety is a relatively stable proneness [51] while state anxiety measures are subject to change as children grow up. van Steensel and Heeman [10] contend that as children with ASD get older, they encounter higher social expectations to adapt to their surroundings and consequently become more aware of their own social difficulties, leading to more stress and anxiety. Considering that state anxiety was measured after the first group intervention session during which participants were introduced to and asked to interact with unfamiliar peers, the state anxiety score could have reflected the participants’ individual variations in cognitive and affective responses to the socially demanding situations. It has been previously suggested that unfamiliarity of the intervention research situations may increase the self-report state anxiety of participants [52]. In this case, the state anxiety of older participants, who were likely to be more aware of the social demands during the intervention session than younger participants, may have been particularly heightened. As the validity and reliability of the STAI have been established under relatively neutral, non-anxiety provoking contexts in previous studies, future studies should explore how the reliability of the STAIS may vary depending on the context of the situation in which the measure is taken.

With respect to the parent-report anxiety, parents may expect their older children to manage their anxiety more effectively and, therefore, show less anxious behaviors. This increased expectation may lead the parents to over-rate the severity of their older children’s anxiety even if the children are in fact showing the same level of anxiety regardless of their age. At the same time, it is also possible that older children do experience more anxiety, and parents are accurately reporting their children’s anxiety levels.

Unlike the findings from van Steensel and Heeman [10] and contrary to our hypothesis, IQ was not associated with any of the anxiety variables in this study. Although we included participants with a relatively wide IQ range of 68-136, we cannot rule out the possibility that different association patterns may emerge when more participants with lower IQ levels (e.g., <70) are included.

All three anxiety variables were associated with the total RRB scores measured by the SRS, but not with the total algorithm score of the ADI-R, partially confirming our hypothesis. We speculate that the total RRB scores of the ADI-R, which are measured by six items, may not have been enough to capture items that are particularly more linked with anxiety. For instance, we found items related to sensory sensitivity to be significantly linked with state anxiety. However, according to the ADI-R diagnostic algorithm, a participant’s score is based on the selection of a higher rating between “interests in part and/or non-functional parts” and “sensory oversensitivity.” If one scored higher in the former question, their total RRB domain score would not have captured constructs related to sensory sensitivity. This calls for the need to use instruments that target specific and nuanced constructs within RRBs in future research. For instance, anxiety in ASD was reported to be associated with specific constructs within RRBs such as insistence on sameness1 and intolerance to uncertainty (IU) [53,54]. Future researchers should conduct follow-up studies with instruments that can validly and directly measure these constructs and examine how different aspects of RRBs are associated with state and trait anxiety.

The significant associations between self-reported trait and state anxiety and the Compulsion/Adherence subscale of the ADI-R mirror the results of Hallet et al. [55], who found that the obsessive-compulsive symptoms of ASD were associated with anxiety for participants with IQ >70. Gillott et al. [56] regarded anxiety as both the source and the consequence of the ritualistic and compulsive behaviors of ASD. That is, individuals with ASD may engage in ritualistic or compulsive behaviors to relieve their anxiety [57] and may also become anxious because social environments are not organized to accommodate their compulsive and ritualistic behaviors.

In terms of the significant associations between state anxiety and parent-reported trait anxiety reported in this study, individuals with high sensory sensitivity may be constantly bothered by various sensory stimuli in the environment and may attempt to regulate the negative, overwhelming emotions with maladaptive strategies such as avoidance or withdrawal [58,59]. Trying to avoid various sensory stimuli may require individuals to be alert to their environment and ultimately lead to an increased likelihood of state anxiety [60-62]. Green et al. [63] also found that sensory oversensitivity predicts anxiety symptoms one year later. Meanwhile, Wigham et al. [19] suggested that IU was a mediator between sensory under- and over-responsiveness and anxiety in children with ASD. Individuals with ASD with high IU may find facing and interpreting aversive and unpredictable sensory stimuli to be stressful and anxious [53,54].

State anxiety was associated with all problem behaviors except for rule-breaking behaviors. Youths with ASD who experience high state anxiety during stressful situations may be engaging in problem behaviors as maladaptive strategies to resolve the anxiety at the moment [22]. It is also possible that youths with ASD who engage in frequent problem behaviors may end up in anxiety-provoking situations and, therefore, stay in higher levels of state anxiety more often than youths with fewer problem behaviors. Moreover, contrary to our initial hypotheses, self-reported trait anxiety was not associated with any of the internalizing and externalizing behavior symptoms in youths with ASD, suggesting that self-reported state anxiety during stressful situations may be more strongly implicated with problem behaviors compared to self-reported trait anxiety.

The findings of this study bear several implications for future studies. To our knowledge, this is the first study that examined the associations between the self-reported state anxiety of adolescents with ASD and their individual and behavioral variables. Future studies should complement the findings of this study by comparing the association patterns between when state anxiety is measured by a self-report instrument and when measured by behavioral or physiological measures (e.g., electroencephalogram in Van Hecke et al. [64]; electrodermal activity in Jain et al. [65]; and heart rate in Mertens et al. [29]).

Except for the Rule-Breaking and Compulsion/Adherence, parent-report trait anxiety was associated with the same variables with which self-report state anxiety variables were associated, while the parent-reported and self-reported trait anxiety measures demonstrated markedly different patterns of associations with the purported variables of interest. These patterns may have emerged because parents are usually making judgments about their children’s trait anxiety levels based on how youths express their anxiety in taxing situations, and these may be better captured in self-report state anxiety items (e.g., “I feel jittery”) than in trait anxiety items (e.g., “I am a steady person”). Additionally, given the high prevalence rate of alexithymia (i.e., difficulty in identifying and describing emotions) in ASD [66], we speculate that the self-reported trait anxiety, which requires objective and long-term self-reflection and awareness of emotions and moods, may not be as accurate as the parent-reported trait anxiety. Moreover, the association patterns of parent-reported trait anxiety may be similar to those of self-reported state anxiety because reporting how anxious they feel at the moment may be easier for adolescents with ASD than reporting on their general proneness.

Previous studies have not reached a consensus on whether parent-report can accurately measure the anxiety levels of adolescents and youths with ASD. Some studies contend that parents may not be able to accurately judge their children’s internalizing symptoms because adolescents with ASD may be hiding their emotional distress from their parents [67]. Others argued that their lack of a theory of mind abilities [42] makes it difficult for individuals with ASD to accurately assess their anxious states and questions the reliability of their self-report results. We urge further research on the assessments of anxiety captured by self- and parent-reports, perhaps using qualitative methods (e.g., cognitive interview) to explore in detail the reasons for their responses to each item and how different contextual variables influence the constructs captured in the items. Moreover, we suggest that future studies measuring the adolescents’ trait anxiety should utilize both self- and parent-reports to gather multi-faceted, objective, and accurate information.

Conner et al. [52] showed that parents’ own state and trait anxiety symptoms were significantly associated with parentreported anxiety and problem behaviors of their adolescent children with ASD. Noting that these anxiety levels were measured before the anxiety intervention of the adolescents with ASD, Connor et al. [52] explained that parents with high anxiety levels may have been more anxious themselves due to the timing of the measurement, and this anxiety may have influenced parents’ interpretation of their children’s behaviors. The same mechanism may have also applied to this study. This underscores the need for studies examining the association between parents’ own anxiety in various anxiety-provoking situations and how they report about their child’s anxiety.

While Kim et al. [23] suggest addressing heightened levels of anxiety as an effective way to decrease problem behaviors, it might be more effective to identify mediators of the associations between problem behaviors and self-reported state anxiety rather than those between problem behaviors and self-reported trait anxiety. Relatedly, targeted interventions training youths with ASD with high self-reported state anxiety to apply adaptive emotion regulation strategies (e.g., cognitive reappraisal) in stressful situations may be more effective than targeting individuals with high self-reported trait anxiety. However, as we cannot draw any temporal or causal conclusions in this study, we should not rule out the possibility that exhibiting problem behaviors put youths with ASD in situations that heighten their state anxiety. Future studies should use a controlled experimental design to investigate the causal relationships between state anxiety and different problem behaviors in individuals with ASD to inform interventions that support the cognitive-emotional and behavioral development of youths with ASD.

There are several limitations to be considered when interpreting the findings of this study. First, the majority of participants had subclinical levels of anxiety, at least as shown in their STAI scores. Although there is no absolute cut-off score for the STAI, typically, individuals with anxiety disorder would score in the range of 47-61 in the STAIT [68] while the mean STAIT score for the participants in this study was 38.05. Second, we could not conduct analyses of the effects of gender because we had only five female participants. Finally, as a correlational study, we cannot make claims about causal relationships between anxiety and behavioral variables.

Self-reported state and trait anxiety and parent-reported trait anxiety were each associated with different sets of demographic and behavioral variables, while self-reported state anxiety and parent-reported trait anxiety showed relatively similar association patterns. Emphasizing that self-reported state and trait anxiety and parent-reported trait anxiety are operating under distinct underlying mechanisms, we urge future researchers who study anxiety of adolescents with ASD to be aware of the instruments’ differences. The findings also highlight the need for future studies exploring precisely what types of anxiety constructs are captured by items in each instrument. Finally, more studies investigating the causal associations between state anxiety and problem behaviors and mediators or moderators of the associations are needed to inform the development of effective cognitive-emotional and behavioral interventions that teach adaptive strategies to youths with ASD.