Paper: "Self-reported sex differences in high-functioning adults
with autism: a meta-analysis"
This paper was published in 2018, in the journal Molecular Autism.
You can click HERE to download a PDF of the paper.
Keep reading to see a plain English summary, or you can watch my explanatory video!
Author's note: I regret the problematic title and language in the paper, which reflects our understanding at the time it was written (2016-2017). The authors were writing with the best intentions, but we now know that functioning labels, person-first and pathologizing language are stigmatizing and hence harmful to autistic people. Please see this page (link opens in new tab) to understand why.
Here is a video summary of this paper. You can open it up to large screen, and turn captions on and off by clicking the 'CC' button.
Why is this an important issue?
Autism has historically been overlooked, under-researched and under-diagnosed in people assigned female at birth (AFAB). One reason for this is that diagnostic instruments and screening tools were designed based on observations of people who were assigned male at birth (AMAB) and assumed to be male by gender.
We now know that autistic features can look different in girls, women and people AFAB. At the time we were working on this paper (2016-2017), however, research on sex differences had mainly focused on children - very little research existed around how autism might look different in adults with different sex and gender identities. This was important because the literature suggested that very often, autistic women and people AFAB were not being recognised as autistic and therefore not receiving any support.
In the UK at the time, screening tools played a very important role in helping to identify those who might be autistic so that they could be assessed and diagnosed. Problematically, very little was known about whether the screening tools we had were appropriate for autistic girls, women and AFAB people. If not, then it would go some way towards explaining why these people were more likely to be overlooked.
What was the purpose of this study, and what did the researchers do?
We looked at one popular screening instrument for autism, the Ritvo Autism Asperger Diagnostic Scale-Revised (RAADS-R). We had obtained RAADS-R data from our local clinic, and collected together more RAADS-R data from researchers who were kindly willing to let us use their data. This kind of study is what is called a meta-analysis: it is when you collect data from across many studies and re-analyse it yourself. We ended up with nearly 1000 datasets, which we divided into four groups: autistic men, autistic women, non-autistic men, and non-autistic women.
We compared these groups on several different aspects of behaviour - the RAADS-R subscales of “social relatedness”, “circumscribed interests”, “language” and “sensorimotor abnormalities”. Participants had self-reported their answers to each scale.
We compared participants in two different ways:
Comparing people by sex, we compared people we assumed to be men and people we assumed to be female, ignoring their neurotype: we did this to see if there were ways that autistic and non-autistic people of the same sex were similar to each other and different to the other “sex”.
Comparing people by diagnosis, we compared autistic people, whatever their sex, to non-autistic people: we did this to see if there were ways autistic people of any sex were similar to each other, and ways that non-autistic people of any sex were similar to each other.
We also looked for places where we might see what are called statistical interactions between our two comparison variables (sex and diagnosis): this would reflect that one group stood out from all the rest. For instance, it might be that while non-autistic men and women were very similar in something, autistic men and autistic women were very different, or vice versa.
What were the results of the study?
We found differences based on diagnosis on all of the subscales of the RAADS-R. This makes sense - it is a screening tool for autism, so you would hope that it could detect differences between autistic and non-autistic people in each subscale! In the “Social relatedness” domain, for instance, autistic people scored more highly than non-autistic people regardless of their sex.
We did find a couple of differences based on sex: regardless of whether they were autistic or non-autistic, women reported less “circumscribed interests” than did men. This fits with what we know about areas of focused interest in autistic people: the intense interests of female or AFAB autistic people are often more “relational” (social, e.g. about relationships or other people) and gender-normative than those of autistic people who are male or AMAB. This is one reason why female and AFAB people are overlooked in diagnostic assessments. It therefore makes sense that female participants, whether autistic or not, did not score highly on the “circumscribed interests” described in the RAADS-R - since the measure was designed based on AMAB people, it is more likely to include interests more typical of these individuals!
Another effect of sex was that regardless of whether they were autistic or not, women reported less neurodivergent language features than did men. A statistical interaction between sex and diagnosis, in language features, reflected that non-autistic men and women were quite different to one another in their use of language, while autistic men and women were very similar to one another. This speaks in favour of the RAADS-R: non-autistic women are often presumed to be more fluent communicators than non-autistic men, so it makes sense that this came across in their scores. In contrast, the RAADS-R seemed to be picking up neurodivergent language features (like struggling with non-literal language) in all autistic participants.
In the sensorimotor domain, autistic people reported more sensory and motor differences than non-autistic people. Interestingly, however, a statistical interaction between sex and diagnosis showed that autistic women reported higher rates of sensory and motor difficulties than did autistic men. This finding was very interesting to us, as sensorimotor features are neglected in the diagnostic criteria for autism and in diagnostic tools. If these features are especially prominent for AFAB and female autistics, they may be under-detected.
What are potential weaknesses in the study?
In 2016-2017 (when the study was written), we were sadly very unaware of the stigma associated with functioning labels, person-first and pathologizing language about autism.
We were also naive about the diversity of sex and gender identities. As we did not collect the data ourselves, we took the descriptions of our participants as “male” or “female” (in the datasets we received) at face value, assuming that they described cisgender people. Because the majority of people ARE cisgender, it’s likely that most individuals were correctly categorised. However, it is likely that our categories of “men” and “women” missed some people who should have been included in these categories, and incorrectly included some people who did not belong in these categories.
Our study of sex differences was restricted to the distinction between (assumed cisgender) men and women. Future research needs to also investigate differences in autistic features between cisgender, non-binary and transgender people.
We did not have a lot of information about our participants, so we did not consider a lot of personal characteristics that might have influenced our findings. For instance, race, ethnicity and socioeconomic status all affect the ways people behave. The RAADS-R is a Westernised, Anglo-centric test which may not fairly represent people with racial/ethnic identities.
Finally, the RAADS-R is based on self-report, and people are often not greatly accurate or objective when reporting their own behaviour, thoughts and feelings.
How will these findings help autistic adults now or in the future?
We aimed, in this paper, to highlight that the screening tools that were in popular use at the time might not be equally likely to identify autism in people of different sexes and genders. Unfortunately, this problem is very much still present. We know that the dominant diagnostic and screening tools for identifying people as autistic are still those based on how autism looks in AMAB people, and that autistic female and AFAB people are still under-recognised.
Researchers and clinicians need to develop tools that are equally likely to identify autism in people who are AFAB, female-identifying, AND who have non-binary and/or transgender identities - since these individuals are highly overlooked within the autistic community.
We also wanted to highlight that in autism research, there is an inherent problem when researchers look for participants who are diagnosed as autistic. Because the dominant diagnostic and screening tools do not detect autism equally in people of all sexes and genders, there are many, many undiagnosed autistic people out there who are not being involved in research, are not being represented, and are not having any say in issues relevant and important to autistic people.
When we researchers compare thoughts, behaviours, feelings in diagnosed autistic people, thinking that we are looking for differences that might arise from characteristics like sex or gender, we need to be aware that we are likely missing many undiagnosed individuals - since these individuals are not making it into research, studies may be perpetuating the status quo of only recognising autistic presentations that fit the more stereotypical "look" of autism. We’re not sure how to resolve this problem, but it’s extremely important and it is good to see studies beginning to reflect inclusion of self-identifying autistic people.
Thank you for reading!