Discussion

The present study aimed to extend previous research regarding the relationships between autism, anxiety and the mediating influence of IU and SOR. Previous research has mainly focused on children with ASC, with limited research studying autistic traits in anxious individuals, particularly adults. Furthermore, there is a shortage of research focusing on autism in women, resulting in arguments that there is a poorly understood female phenotype which may be characterised by elevated anxiety in women (Ludlow et al., 2015), and thus the argument that women diagnosed with anxiety disorders may include some of the undiagnosed female ASC population (Kim et al., 2011). In line with the first hypothesis results indicated women with anxiety disorder diagnoses had significantly higher autistic traits on average than general population samples, with 21.1% meeting the clinical cut-off (over 20 times higher than general population). This finding may be explained by the camouflaging hypothesis, in that women with autism have higher anxiety due to the stress of masking their autistic traits (Lai et al., 2017). Indeed, there were significant correlations between all measures of anxiety and autistic traits, with each measure also correlating with the two mediating factors of IU and SOR, in line with previous findings (Bitsika et al., 2016).
While previous research has established ASC severity as a reliable predictor of anxiety in autistic individuals (e.g. Lever & Guerts, 2016; Murry et al., 2019), little research has investigated whether anxiety can predict autistic traits. The present study found all four anxiety subtypes were individually significant predictors of autistic traits, supporting the findings of van Steensel et al., (2013) in children. Additionally, all four anxiety subtypes were predictive of both IU and SOR, with these two factors also being significant predictors of autistic traits.
Finally, IU has been identified as a key construct in a range of anxiety disorders (Norr et al., 2013) as well as demonstrating links with ASC (Neil et al., 2016), thus South and Rodgers (2017) propose that IU is one mediator between ASC and anxiety, however previous research has focused on autistic samples rather than those with anxiety disorders. As far as we are aware, the present study has been the first to demonstrate that IU was a significant partial mediator between all four anxiety subtypes and autistic traits. For all anxiety subtypes the effect of anxiety symptoms was not completely reduced when accounting for IU, suggesting partial mediation in the model and that other factors may be of influence. Indeed, SOR was also found to be a significant partial mediator between all anxiety subtypes and autistic traits. The present findings were in line with South and Rodger’s (2017) model and with research linking SOR to a range of anxiety subtypes (Bitsika et al., 2016).
As noted, IU and SOR only partially explained the relationship between anxiety and autistic traits, indicating the presence of other mediating factors. Indeed, South and Rodgers (2017) have identified alexithymia as a further possible mediating factor that has been shown to mediate the relationship between autistic features and anxiety (Morie et al., 2019) though research has yet to investigate a possible mediating relationship in the opposite direction. Additionally, alexithymia has also been associated with SOR, bringing up another area of possible further research regarding investigating the relationships between the mediating factors themselves. Indeed, IU mediated the relationship between sensory features and anxiety (Uljarević et al., 2016), and in autistic children but not typically developing controls, SOR was found to mediate the relationship between both separation anxiety and Specific Phobia symptoms and the autistic trait of insistence on sameness (Black et al., 2017).
Further variables that could also be investigated for potential mediating effects include RRBs, which have been associated with both ASC and OCD (Stewart et al., 2016) and have been linked to increased anxiety in autistic samples (Rodgers et al., 2012). Indeed, Wigham et al. (2015) found that the relationship between sensory atypicalities and RRBs was mediated by IU, though research has yet to investigate the role of RRBs in elevated autistic traits in anxious samples. Even within the mediating variables investigated, there are further areas that could be researched. For instance, both sensory over- and under-responsivity has been associated with ASC (Ben-Sasson et al., 2007), as well as a range of atypical processing across different domains of sensory function including auditory and visual (Tavassoli et al., 2014), each of which may have differential relationships with ASC and anxiety. For instance, Bitsika, Arnold & Sharpley (2019) found that auditory-specific sensory behaviours shared the strongest correlation with parent-rated anxiety in autistic children and also identified the further variable of sensory avoiding, finding that ASC symptoms influenced elevated levels of parent-rated GAD indirectly through greater levels of sensory avoiding.
Understanding the relationship between autism and anxiety has significant clinical implications, not only for the diagnosis of both disorders but for the treatment of the two as comorbidities (Sung et al., 2011). Anxiety in ASC is associated with increased levels of repetitive behaviours (Lidstone et al., 2013), externalising problems including aggression and self-injury (Mattila et al., 2010) and loneliness (White & Roberson-Nay, 2009). Using the ASC-ASD anxiety measure in autistic children, those scoring above the cut-off had poorer scores on a range of quality of life (QoL) domains including social, emotional, physical and school functioning, with regression analyses demonstrating that the items relating to IU in particular predicted QoL domains as well as poorer QoL overall (Adams, Clark & Keen, 2019), emphasising the need for ASC adapted treatment (Rodgers & Olfield, 2018). Furthermore, as multiple anxiety disorders often occur concurrently in ASC, symptoms are likely to have both interaction and additive impacts on treatment efficacy. Adaptations to CBT for the treatment of those with ASC include being taught social skills to recognise emotions and identify triggers for anxiety (Sung et al., 2011), parent management techniques to address broader deficits that may be contributing to anxiety (Moree & Davis, 2010) and using virtual reality to aid exposure treatment for phobias (Maskey et al. 2019), the latter being useful for circumventing the common difficulties with imagination in ASC (APA, 2013) which may impair engagement in imaginal desensitisation. Further factors that have been identified as important for the treatment of autistic individuals include regular sessions in a structured setting, consistent therapists and social exposure as well as the use of autism-friendly strategies such as visual strategies or role-plays (Sung et al., 2011).
The results of the present study suggests that targeting IU and SOR may be effective for treating anxiety, and indeed Keefer et al., (2017) found that pre-intervention IU predicted higher anxiety in both pre- and post-intervention measures, therefore indicating that directly targeting IU may improve outcomes. Such treatments are in development, with Rodgers et al., (2016) devising the Coping with Uncertainty in Everyday Situations Program focused on IU, designed to specifically address a target uncertain situation identified by the patient, and has shown promise in both child (Rodgers et al., 2016) and adult samples (Rodgers, Herrema, Honey & Freeston, 2018).
Regarding SOR as a target in the treatment of anxiety, treatments such as Sensory Integration Therapy have focused on helping those with ASC to regulate sensory experiences in order to alleviate ASC symptoms and anxiety, but evidence is mixed and based primarily on case-study designs which may not be generalizable (Lang et al., 2012), therefore more research is required to investigate SOR as a potential target in the treatment of anxiety. Finally, understanding autism in women is of clinical significance as the majority of women report negative experiences during the diagnostic process (Milner, Mcintosh, Colvert & Happé, 2019) and the lack of understanding is suspected to lead to delayed diagnosis (Bargiela et al., 2016). Reducing this delay is significant when those who are not diagnosed until adulthood are less likely to receive adequate educational support compared to those diagnosed earlier (Baldwin & Costley, 2016).
Despite being one of the few studies to investigate autistic traits in anxious samples, there are a number of limitations to this study, including that the relationship between autistic traits and anxiety symptoms may be a consequence of diagnostic overshadowing between ASC and anxiety disorders due to overlapping diagnostic criteria. For example, SAD and ASC have a number of shared symptoms, including social avoidance and a failure to speak in social situations (APA, 2013), with 57.1% of an autistic sample reporting clinically significant SAD symptoms (Kuusikko et al., 2008). White et al. (2012) argue SAD and ASC are separable but highly correlated disorders, with unique factors relating to each disorder, and evidence supports different developmental trajectories of shared deficits in emotion recognition, which are traced to early delays in mimicry and imitation in ASC yet have been related to temperamental and environmental factors such as developing biases towards threatening interpretations of stimuli in SAD (Tyson & Cruess, 2001). Cath, Ran, Smit, Van Balkom and Comijs (2008) found that the AQ subscales of communication problems and lack of imagination discriminated between those with a dual diagnosis of ASC and SAD and those only diagnosed with SAD. Similarly, Hartley and Sikora (2009) found that scores on communication and social interaction difficulties scales could differentiate between those with ASC and those with anxiety disorders. Finally, while White, Schry & Kreiser (2014) acknowledge that distinguishing the symptoms of SAD from those of ASC may be difficult for clinicians, they conclude that the two are separable clinical constructs, and that social anxiety can be statistically separated from ASC symptom severity (White et al., 2012).
As well as possible conceptual limitations, there are a number of methodological limitations, including issues regarding how mediation might not be best tested cross-sectionally, (Cole and Maxwell, 2003). Future research should aim to explore these findings using a longitudinal design.
A further limitation relates to the anxiety measures used, the majority of which are designed for use in typically developing individuals and thus may not accurately differentiate autism and anxiety symptoms. This is particularly true when considering that emotion recognition deficits associated with autism, such as alexithymia, may result in difficulties in accurately self-reporting anxiety (White et al., 2014).
In conclusion, the present study has demonstrated that autistic traits can be predicted by anxiety, and that this relationship is partially mediated by SOR and IU. However, there is a lack of research on autism in anxious populations and more research is required to corroborate these findings, as well as investigating other factors which may also play a role in the relationship between autism and anxiety, such as alexithymia. Further research would also benefit from tackling some of the limitations of this study, for example by using a longitudinal design, and exploring these findings in other populations.