AQ items are summed to give a total score, with suggested cut-offs for potentially clinically significant ASC. The scale measures two main higher order factors that related to represent areas describing either difficulties or core traits that are typically present in people with ASC; hence difficulties in social behaviours and a fascination with number and patterns. The first factor is further divided into subdomains relating to social skills and communication, attention switching and to detail, routine behaviours, and imagination.
Therefore, the existing evidence suggests that the high rates of autistic traits found in studies including adult transgender people appears to apply to those assigned a female sex at birth (Jones et al. 2012
). This may be explained by the extreme male brain theory (EMB) (Baron-Cohen 2002
), which suggests that a higher number of autistic traits is associated with exposure to higher levels of foetal testosterone (Auyeung et al. 2009
; Baron-Cohen et al. 2015
). Affected individuals may therefore have more traits associated with a ‘male brain’ such as an increased drive to systemize and lower levels of empathy (Baron-Cohen 2003
). Only one study has used the AQ to investigate which specific autistic traits are particularly prominent in transgender populations (Pasterski et al. 2014
). This study did not exclude participants who had received cross-sex hormone therapy before their first assessment at the transgender health service. As cross-sex hormone therapy reduces anxiety this may have a direct effect on social skills (e.g. increase in social confidence and social functioning, experiencing fewer problems with socialisation and less social distress), one of elevated autistic traits (Bouman et al. 2016a
; Gómez-Gil et al. 2012
). Other limitations in the small existing literature are lack of an adequate cisgender comparison group and small, poorly matched samples.
Clinical reports and previous studies have suggested the existence of high levels of autistic traits among transgender people attending clinical services seeking gender affirming medical interventions (Glidden et al. 2016
). Although previous studies have investigated levels of ASD in children attending gender identity services (de Vries et al. 2010
) very few have focused on the adult population. Those studies recruiting adult transgender people, whether via clinical services or online surveys, have reported high levels of autistic traits and diagnoses (Kristensen and Broome 2016
; Pasterski et al. 2014
). However, those studies have failed to compare their results with a matched controlled population of cisgender people. The need for matching in this field is particularly important as autistic traits are higher on average amongst males (Baron-Cohen et al. 2001
; Baron-Cohen 2002
). Therefore, with the aim of investigating the levels of autistic traits in the transgender population attending services and whether those traits were different to a matched cisgender group, this study recruited a large population of treatment-seeking transgender adults and compared them to a large cisgender control group.
The study found that when comparing both groups there were no significant differences in the number of people who presented with scores suggesting a possible diagnosis of ASC. In fact, the levels of possible caseness of ASC was found to be high in both groups, with nearly a third of participants in both the transgender (36.3%) and the cisgender (33.2%) groups scoring above the clinical cut-off for possible ASC caseness. The findings of this study suggest higher prevalence rate of possible ASD cases among transgender individuals than some studies (Pasterski et al. 2014
), but lower than others (Jones et al. 2012
; Kristensen and Broome 2016
). The sampling techniques and different cut-offs used by previous studies make direct comparisons difficult. However, the general clinical impression of an over-representation of people displaying autistic traits among transgender clinical populations when compared to cisgender people from the general population (Glidden et al. 2016
) was not confirmed by the findings of the present study.
This study confirmed previous findings from adult studies that there is no evidence of increased rates of autism in transgender populations as a whole. It should be noted that the number of possible ASC caseness in both cisgender and transgender people is significantly higher than the prevalence of ASC which has been estimated at 1.1% in the general population (Brugha et al. 2011
). Thus, the large number of cisgender people with possible ASC may have affected the results of the study. A possible explanation may be that the AQ demonstrates a stronger negative predictive value for ASC than a positive predictive value in the samples. The AQ being a screening tool rather than a diagnostic tool fits with this interpretation. Studies investigating levels of ASD in the general population have found similar rates of clinical ASD caseness (Kristensen and Broome 2016
). Hoekstra et al. (2011
) developed and validated the AQ-short and found that both UK and Dutch control samples reported average AQ-short total scores of approximately five points lower than our cisgender sample.
It was when exploring the results according to gender that the findings become more significant. Although no differences were found between people assigned male at birth (possible transgender female and/or non-binary people) and cisgender males, a significant difference was found when comparing people assigned female at birth. The study found that within people assigned female at birth the transgender group (possible transgender males and/or non-binary people) was about twice as likely to have clinically significant levels of autistic traits compared to cisgender females. This is in keeping with the findings of previous research using different versions of the same measure (AQ) (Jones et al. 2012
; Kristensen and Broome 2016
; Pasterski et al. 2014
). With regard to the cisgender sample the findings of the present study appear to support previous research suggesting the presence of a male bias in the presentation of autistic traits (i.e., EMB theory) (Baron-Cohen et al. 2001
; Baron-Cohen 2002
; Hoekstra et al. 2008
With regard to the transgender sample, the present results suggesting that transgender assigned females at birth suffer from higher overall autistic traits than transgender assigned males at birth are partially in line with findings of previous research (Kristensen and Broome 2016
; Jones et al. 2012
). A possible explanation for this might still be the abovementioned EMB theory (Baron-Cohen et al. 2001
), which posits that assigned males at birth are more likely to and have the tendency to suffer from higher levels of autism. Consequently, transgender men might also face more social difficulties and worries related to having to change gender-roles and thus gender-specific behaviours (e.g. not feeling masculine enough, fear of being perceived as female), which might lead to experiencing higher levels of social anxiety. These results are in conflict with past literature suggesting higher levels of social anxiety in transgender women due to the reduced social acceptance of gender variant behaviours amongst assigned males than assigned females (e.g. society being more accepting of masculinity than femininity) (Yu et al. 2017
). Future research should try to properly understand the different impact of autistic traits onto transgender assigned females as well as on assigned males at birth.
There is a feeling amongst clinicians that current screening and diagnostic tools may be biased towards identifying and diagnosing assigned males. As a consequence, assigned females have been under diagnosed by current clinical practice (Gould and Ashton-Smith 2011
; Lai et al. 2011
; Brugha et al. 2016
The present study also examined AQ-short subscale scores to understand differences in autistic traits between groups. The study found that there were differences in the autistic traits between groups. Transgender people (particularly those assigned female at birth) were found to have specific difficulties in relation to social behaviours, including social skills, mental flexibility and problems switching attention, the disruption of which could lead to increments in anxiety levels (Batten 2005
). These findings could be associated with social anxiety, which is known to be highly prevalent among the transgender population (Millet et al. 2017
). The transgender group, however, reported fewer difficulties in relation to other autistic traits of poor imagination, and fascination with numbers and patterns. Therefore, there may be a confounding element where the transgender sample scored higher on the AQ without the presence of ASC. This may indicate that some transgender people do not really present with ASC but the high levels of social difficulties due to anxiety, depression and years of victimisation may affect the way they interact with others, as suggested by Turban and van Schalkwyk (2018
). In fact interpersonal difficulties among this population have already been described (Davey et al. 2016
). However, this would not explain higher scores on attention switching or attention to detail.
Enhanced scores on the subscales for routine and attention switching in the transgender population might be linked to the elevated social anxiety symptomatology experienced by this population. In fact, previous research has suggested that autistic individuals report similar scores on the attention switching subscale of the AQ-50 to people with a diagnosis of social anxiety disorder (Cath et al. 2008
). They also stated that socially anxious people have been found to display higher scores than non-clinical controls (Cath et al. 2008
). Thus, high scores on the routine subscale might reflect an increase in social anxiety levels due to the social difficulties related to changes in routine behaviours (e.g. switching gender roles, which imply learning and practicing different gender-specific behaviours), whilst high scores on the attention switching subscale might indicate difficulties in altering the focus of attention from negative and anxiety-producing experiences (e.g. being looked at in the street might lead to an increase in social anxiety levels due to the transgender individual’s fear of being recognised as trans) to more positive and adaptive coping behaviours (e.g. taking into consideration that being looked at in the street might be due to reasons other than the person being recognised as transgender). Transgender people are known to face these social challenges during the time of transition (Grossman and D’augelli 2006
), which might result in them scoring higher than cisgender people on the subscales related to such difficulties.
As suggested by Lombardi (2001
), transgender individuals’ extreme marginalisation and vulnerability might lead them to experience increased psychological, health and especially social isolation, when compared to other social groups. Consequently, some transgender people might score particularly high on some AQ-short items and subscales because of their dysphoria and related anxiety as well as negative past experiences (e.g. transphobia, bullying) (Skagerberg et al. 2015
). Thus, certain scores on the AQ-short might simply indicate increased social difficulties, which would increase their AQ-short total scores reaching an ASD caseness level, while not being driven by the presence of ASC. This idea is supported by literature investigating autistic traits in socially anxious populations with the use of both AQ-50 (White et al. 2012
) and AQ-short (Tonge et al. 2016
). These studies suggested that the scores in the subscales related to social functioning and behaviours might be inflated by the difficulties experienced in social interactions and that scores on the AQ subscales related to social functioning need to be interpreted with caution when the subjects display marked social difficulties (Tonge et al. 2016
; White et al. 2012
). Against this, studies recruiting those with social anxiety may well be sampling individuals who have higher levels of autistic traits. It will be important to try to disentangle these in the future.
Notwithstanding the strengths of the present research (e.g., large sample size, matched controls, and homogeneity of sample in terms of treatment status), there are several limitations. First, as a case–control study was adopted participants should have been matched on additional variables other than age and sex assigned at birth (e.g., IQ, educational level) to properly eliminate the issue of confounding and gaining appropriate efficiency. Second, as previously discussed the AQ-short is a self-reported assessment tool for autistic traits instead of a diagnostic tool. Third, the data relative to the cisgender participants was gained through the “Cambridge Psychology” website and snowballing sampling and response bias might have biased the results as high levels of autistic traits were found in this population. Fourth, transgender and cisgender groups completed different versions of the AQ, which might have had an impact on how the participants approached and answered the items of the scale, although the items collected from the AQ-50 were the same as per the AQ-short.
Future studies may focus on analysing the role of gender affirming treatment (cross-sex hormones) in autistic traits as measured by the AQ. It has been noted clinically that transgender people with a previous diagnosis of ASC express fewer and less obvious autistic traits following medical transition. This has raised questions about the diagnostic safety of an ASC diagnosis made in transgender people. Studies which include diagnostic tools such as a clinical interview and/or formal diagnostic tools such as the Autism Diagnostic Interview Revised (ADI-R—Lord et al. 1994
), Autism Diagnostic Observation Schedule (ADOS—Lord et al. 2002
) and Diagnostic Interview for Social and Communication Disorders (DISCO—Wing et al. 2002
) and which control for transitional status and anxiety are needed to fully explore the nature of elevated autistic traits in transgender populations.
Overall, this study found that autistic traits appear to be more prevalent in transgender people assigned female at birth, but not in those assigned male at birth. The autistic traits found in both groups appear to be particularly connected to social skills, which may be associated to the high levels of anxiety and low self-esteem that this group often experiences (Claes et al. 2015
; Bouman et al. 2016b
; Millet et al. 2017
). As transgender people displayed issues with the social features of ASC whilst reduced difficulties with autistic traits that may be less likely to be mimicked by the experience of gender dysphoria, undergoing gender affirming medical treatment might have a positive impact by improving interpersonal skills and social interactions. Longitudinal studies investigating the role of cross-sex hormone treatment on autistic traits are therefore needed as they may help reaching more robust evidence as to whether the observed autistic traits equate to clinical ASC or not.