BBC’s “Transgender Kids, Who Knows Best?” p1: You got Autism in my Gender Dysphoria!

This series on BBC’s “Transgender Kids: Who Knows Best?” is co-authored by HJ Hornbeck and Siobhan O’Leary. It attempts to fact-check and explore the many claims of the documentary concerning gender variant youth. You can follow the rest of the series here:

  1. Part One: You got Autism in my Gender Dysphoria!
  2. Part Two: Say it with me now
  3. Part Three: My old friend, eighty percent
  4. Part Four: Dirty Sexy Brains


Petitions seem as common as pennies, but this one stood out to me (emphasis in original).

The BBC is set to broadcast a documentary on BBC Two on the 12th January 2017 at 9pm called ‘Transgender Kids: Who Knows Best?‘. The documentary is based on the controversial views of Dr. Kenneth Zucker, who believes that Gender Dysphoria in children should be treated as a mental health issue.

In simpler terms, Dr. Zucker thinks that being/querying being Transgender as a child is not valid, and should be classed as a mental health issue. […]

To clarify, this petition is not to stop this program for being broadcast entirely; however no transgender experts in the UK have watched over this program, which potentially may have a transphobic undertone. We simply don’t know what to expect from the program, however from his history and the synopsis available online, we can make an educated guess that it won’t be in support of Transgender Rights for Children.

That last paragraph is striking; who makes a documentary about a group of people without consulting experts, let alone gets it aired on national TV? It helps explain why a petition over something that hadn’t happened yet earned 11,000+ signatures.

Now if you’ve checked your watch, you’ve probably noticed the documentary came and went. I’ve been keeping an eye out for reviews, and they fall into two camps: enthusiastic support

So it’s a good thing BBC didn’t listen to those claiming this documentary shouldn’t have run. As it turns out, it’s an informative, sophisticated, and generally fair treatment of an incredibly complex and fraught subject.

… and enthusiastic opposition

The show seems to have been designed to cause maximum harm to #trans children and their families. I can hardly begin to tackle here the number of areas in which the show was inaccurate, misleading, demonising, damaging and plain false.

… but I have yet to see someone do an in-depth analysis of the claims made in this specific documentary. So Siobhan is doing precisely that, in a series of blog posts.

So why am I tagging along? Two reasons:

  1. I bring a useful set of skills. Between my many blog posts on statistics and science, and my lectures on gender and sexism, I can speak with some authority on certain things.
  2. There’s this bizarre tendency to dismiss the input of transgender people as the work of “transgender activists.” This is fractally wrong: is someone who advocates for themselves an “activist,” even if they’re part of no organization? Since when did “activists” become the baddies? And how does being gender dysphoric disqualify you from having an understanding of gender dysphoria?!When cis het white guys start agreeing on the facts too, “activism” magically becomes “advocacy.” It’s one of our superpowers.

Who Cares?

I’ll get us warmed up, with a discussion of the link between gender dysphoria and autism. The documentary mentions it in passing:

NARRATOR: There is also evidence of a link between gender dysphoria and autism. One study found that children with gender dysphoria are seven times more likely to be on the autistic spectrum than children from the general population.

DR. KEN ZUCKER: It’s possible that kids who have a tendency to get obsessed, or fixated on something, may latch on to gender.

Lest you think this was just some fringe belief, here’s three mainstream activist groups buying into it:

While research suggests a large overlap between transgender and autistic communities, trans autistic people often lack access to services and supports that understand and respect all aspects of their identity.

Our first step should be to look at the bigger picture: why should we care? What does it mean if autism and gender dysphoria are linked?

As far as I can see, jack shit. Imagine if people who got high scores on math tests were more likely to have brown eyes. Would this mean brown corneas are intrinsically better for reading math? Ridiculous. Maybe there’s a shared genetic cause behind the scenes? That’s actually reasonable, but it only matters on a personal level if you can change your genes. Policy wonks are free to weigh the pros and cons of using eye colour as a proxy for math scores, but if a blue-eyed person did great on a math test you wouldn’t accuse them of cheating.

Does knowing of this correlation change how you’d treat someone with low math test scores? It better not. People who are struggling deserve help, especially if the cause of those struggles come from no fault of their own.

While test scores are quite different from dysphoria, the broad strokes still cover the canvas. This association should not change how we treat autistic-spectrum or trans* people on a personal level; if it does, that implies we’ve assigned a moral judgment to a condition to a person, based on something they had thrust on them. Or, as someone else more eloquently put it,

The best option is to allow someone to explore their feelings, support them in gaining self-understanding, and accept their identity whatever it turns out to be. It is not complicated, and it’s only scary if you are still holding onto the belief that being either autistic or transgender – or, perish the thought, both – is a terrible thing to be. Which it’s not. I am, along with countless others like me, living proof of that.

(Incidentally, they also agree that there’s a correlation between autism and gender dysphoria. Told you it was common).

Let’s Science the Shit out of This

But ignore basic rationality for a second. How did this link between autism and gender dysphoria get uncovered? Through SCIENCE, of course! In this case, the bulk of the literature is case reports: a doctor or councilor would notice a number of trans* people with autism diagnoses, and write up a report. Here’s the earliest I can find:

Abelson, A. G. “The Development of Gender Identity in the Autistic Child.” Child: Care, Health and Development 7, no. 6 (November 1, 1981): 347–56. doi:10.1111/j.1365-2214.1981.tb00851.x.

The flaw should be obvious: the plural of “case report” is not “prevalence.” The former is a special case, by definition, that does not form a representative sample. To assess the latter, we need to study all cases and remove bias.

Fortunately, it only took a few decades for those studies to start arriving. Here’s a highly-ranked paper on the subject.

Jones, Rebecca M., Sally Wheelwright, Krista Farrell, Emma Martin, Richard Green, Domenico Di Ceglie, and Simon Baron-Cohen. “Brief Report: Female-To-Male Transsexual People and Autistic Traits.” Journal of Autism and Developmental Disorders 42, no. 2 (February 2012): 301–6. doi:10.1007/s10803-011-1227-8.

Right from the abstract, we can tell something’s up. It does not claim a correlation between gender dysphoria and autism! Instead, we learn that “Transmen had a higher mean AQ than typical females, typical males and transwomen, but lower than individuals with AS.” Catch that? The AQ or Autism-spectrum Quotient scores of trans* women were no different from their cis counterparts.

Figure 1, from Jones [2012]Ugh, that chart is ugly.

My version of figure 1 from Jones [2012]Much better, the uncertainty is a lot easier to read. Anyway, this makes it clear that while the trans* men in this study are an outlier, they are still quite some ways from being in the autism spectrum group. Alas, you’d never gather that from the Results section; because of the way p-values work, the researchers only analyzed how different the trans* men group was from the controls. This leads to a distorted picture of difference.

Speaking of which, my own chart is misleading. It only describes the uncertainty of the mean score of each group, it does not show how AQ scores were distributed. Mistake one for the other, and you walk away with the impression that all of these groups are quite precise and have little overlap.

The distribution of AQ scores, pulled from Jones [2012]

Fortunately, Jones [2102] gives us enough data to reconstruct what the scores might have looked like. I scaled the area of each graph to match the samples size, too. Besides the increase in mess, it’s also eyebrow-raising that the gender dysphoric group with the greatest sample size was closer to the control group, while the outlier group had a much smaller relative sample size; maybe this is all just statistical noise? It would explain why the researchers found as many trans* men within the “Medium” diagnostic criteria as the more extreme “Narrow” one. That shouldn’t happen if trans* men’s AQ scores fall into a normal distribution; you’d expect the percentages of each three groups to be 24.6%, 16.4%, and 9.7% respectively, not the 21.3 / 14.8 / 14.8 split found in the paper.

(Incidentally: there’s three diagnostic levels for autism? That smells of diagnostic flexibility, where inconsistent test results are swept under the rug by letting other criteria overrule them. When your test puts 12% of all cis women, 18.6% of trans* women, and 22.4% of cis men in the autism spectrum, either your test isn’t good or autism isn’t much of a disorder.)

That can’t be the end of it, though. The file drawer effect warns us from relying on the findings of a single study, as there’s a systematic bias in what science gets reported that skews the record toward junk science. The best way to compensate is to reach for a meta-analysis, which pools study findings together and throws a bunch of math at them.

Glidden, Derek, Walter Pierre Bouman, Bethany A. Jones, and Jon Arcelus. “Gender Dysphoria and Autism Spectrum Disorder: A Systematic Review of the Literature.” Sexual Medicine Reviews 4, no. 1 (January 2016): 3–14. doi:10.1016/j.sxmr.2015.10.003.

This is not a meta-analysis. No-one has whipped one up yet. Instead, it’s a systematic review: researchers dig up as many studies as they can, weed out the poor ones based on the quality of their methodology, then try to guess at what the math would have told them.

So what did Glidden et al. guess? That ain’t clear.

The studies that have been conducted [in adults] appear to suggest that, in some cases, there is a co-occurrence between symptoms of gender dysphoria and ASD, although the studies are more limited and less robust than those in the child and adolescent literature. […]

The limited literature in this field does not allow for clear conclusions regarding the prevalence or etiology of ASD in trans people. Studies using good diagnostic procedures or tools are usually limited by the small numbers of subjects; however, when the number of cases increases, the validity of the diagnosis of ASD or gender dysphoria is limited.

So there’s enough literature to suggest autism is correlated with gender dysphoria, but not enough to know how many gender dysphoric people have autism? How then do we tell the difference between a legitimate difference in a sub-population and a statistical fluke due to small samples sizes? That last quoted line is worrying, as you’d expect more evidence to lead to firmer conclusions. If the reverse is true, it’s a sign that contradictory studies are getting file-drawer’d and the consensus isn’t as strong as it appears.

In the meantime, we’re still without a consensus. Let’s go digging for another study.

Pasterski, Vickie, Liam Gilligan, and Richard Curtis. “Traits of autism spectrum disorders in adults with gender dysphoria.” Archives of sexual behavior 43.2 (2014): 387-393.

This time there’s no chart in the paper, so it’s all on me.

Table 2, from Pasterski [2013]

Interesting, trans* men are no longer an outlier! The sample size is a lot smaller, alas, hence the greater uncertainty. Still, this seems like decent evidence for a statistical fluke in Jones [2012], and that the correlation is overhyped or non-existent. I’m sure the authors of this paper will agree.

The current report makes two contributions to what we know about the observed co-occurrence of [gender dysphoria] and autistic traits. First, we have shown that the association appears to be genuine and persists into adulthood.

…. Huh?! No, your data doesn’t show that! You even say so yourself:

For overall AQ scores, there was no significant difference between the MtF transsexuals in our sample and control males, while the FtM transsexuals scored slightly higher than control females on the full scale AQ …

This dataset does show that 7.1% of trans* men met the “autism threshold” (a score above 32, which doesn’t map to the three “phenotypes” from before) vs. 3.9% of control men. For women, the numbers are 4.7% vs. 1.0%. This, and not the mean score of each group, must be why Pasterski [2013] concluded they’d confirmed the correlation. Unfortunately, it’s also evidence their dataset isn’t normally distributed, as only 2.7% of trans* men and 1.5% of trans* women should have passed that threshold according to the listed means and standard deviations. Either statistical noise struck again, or they botched their statistical analysis, or the researchers took a biased sample.

This paper also falls into the p-value trap, and only compares trans* people to controls without autism. It goes further though, by splitting their already-small dataset into smaller chunks in search of outliers. This is re-rolling the dice to get a better outcome, which statistical noise will happily deliver if given enough chances.

It’s shoddy work. Who would publish such a thing? To Wikipedia!

The Archives of Sexual Behavior is a peer-reviewed academic journal in sexology. It is the official publication of the International Academy of Sex Research. The journal was established in 1971 by Richard Green, who served as its editor-in-chief until 2001 He was succeeded by Kenneth J. Zucker.

Wait, Kenneth J. Zucker?! He was the editor of that paper? No wonder he’s a fan of the results.

Still, that systematic review claimed the evidence for an autism-gender-dysphoria link was strong in children. The quote which started this all was about kids. Let’s have a boo, then.

Hold up, wait a minute!

But first, hold that thought.

Very few people think of autism as an acute condition, as something you “get over” eventually. Scientists aren’t sure what causes it, but genetic factors play a strong role. Likewise, gender dysphoria can be classified as a “strong and persistent distress with physical sex characteristics or ascribed social gender role that is incongruent with persistent gender identity” that “is clinically significant or causes impairment in social, occupational, or other important areas of functioning and this distress or impairment is not solely due to external prejudice or discrimination.”

If there was a correlation between autism and dysphoria, we’d expect it to be consistent across child- and adulthood. But if I was able to show there’s no evidence for a correlation in adults, then how can there be a strong correlation in childhood? That can only happen if a) either autism or gender dysphoria are not persistent, contrary to how they are defined, or b) one or both conditions are being overdiagnosed in children. Given that children are adults that haven’t fully baked, it should not be a surprise that their behaviors can vary substantially more than adults. This makes it easy to over-diagnose one condition or the other, artificially inflating the correlation.

Won’t Someone think of the Children?

Still, we shouldn’t let pesky things like logic get in the way of the science.

Strang, John F., et al. “Increased gender variance in autism spectrum disorders and attention deficit hyperactivity disorder.” Archives of sexual behavior 43.8 (2014): 1525-1533.

This seems to be the source of Dr. Zucker’s “seven times more likely” claim. This paper looked at a number of groups, such as kids with epilepsy and ADHD, and had two control groups. Tell me if you can spot the odd person out, though.

Participants with an ASD were diagnosed based on expert clinical impression using the DSM-IV-TR. …

Participants With ADHD were diagnosed based on a comprehensive diagnostic assessment which included clinical neuropsychological evaluation, family interview, and teacher informant reports by an expert clinical neuropsychologist with 12 years of experience diagnosing ADHD (L.E.K). …

Children with epilepsy were diagnosed by neurologist based on clinical features, neurologic examination, ictal video-EEG,and high resolution epilepsy protocol MRI. Children with neurofibromatosis 1 (NF1) were diagnosed by neurologist according the National Institute of Health(NIH) standard for diagnosing NFI (Gutmannetal., 1997; Wolkenstein, Freche, Zeller, & Revuz, 1996). ..

The wish to be the other gender was assessed with the [Child Behavior Checklist] for ages 6–18 years (Achenbach & Rescorla, 2001). The CBCL is a standardized parent report questionnaire of child / adolescent behavioral and emotional functioning.

So they brought out the experts for everything but gender dysphoria? And they didn’t ask the kids what they thought of their own gender, but instead queried their parents? To recall prior behavior they’ve seen in their kids, when we’ve known human memory is fallible for decades? When the parent’s beliefs may be coloured by “religious wisdom” or “common knowledge?” To add further salt to the wound, the Child Behavior Checklist wasn’t designed for gender dysphoria.

The Child Behavior Checklist (CBCL; Achenbach & Edelbrock, 1983) measures a broad range of behavioral and emotional problems and is widely used for clinical and research purposes. Achenbach and Edelbrock (1983) conducted a factor analysis to develop 118 items measuring problem behavior for children and adolescents. A result of this study, the CBCL consists of 140 items that assess behavioral or emotional problems of children aged 6 to 18 years.

Only two of its questions could plausibly deal with gender dysphoria. A quick Google Scholar search shows no-one has validated its ability to assess gender dysphoria either; contrast and compare with the Sexual Experiences Survey. This invalidates the paper’s methodology and makes the data moot.

Though it’s not like the data was that hot to begin with. Want to know how that paper concluded autism was seven times more likely in trans* people? They “diagnosed” eight children out of 147 as having gender dysphoria, when their control groups suggested they should have seen either two or zero. No really, their primary control group of 165 children had zero diagnoses of dysphoria. No wonder the p-values were off the charts. A Bayesian analysis is better suited to small datasets, and here it endorses a difference in prevalence between the ASD and control groups…

Gender dysphoria prevalence rates, from Strang [2014]… but one a helluva lot smaller than a frequentist approach gets you.

But again, you can’t just rely on a single paper. So, let’s finish off this post with another highly-cited one.

Vries, Annelou L. C. de, Ilse L. J. Noens, Peggy T. Cohen-Kettenis, Ina A. van Berckelaer-Onnes, and Theo A. Doreleijers. “Autism Spectrum Disorders in Gender Dysphoric Children and Adolescents.” Journal of Autism and Developmental Disorders 40, no. 8 (August 1, 2010): 930–36. doi:10.1007/s10803-010-0935-9.

This paper takes the opposite approach of Strang [2014], by looking at kids with gender dysphoria and counting the number who have autism. Its methodology is light-years ahead of the previous one.

Participants were 231 children and adolescents who had been referred to the Gender Identity Clinic of the VU University Medical Centre in Amsterdam between April 2004 and October 2007. This is the only clinic in the Netherlands with a multidisciplinary team examining gender dysphoric children and adolescents.

Its sample size, on the other hand (emphasis mine)…

According to the DISCO algorithms, seven (six boys, one girl) of the 11 ASD suspected children had ASD. The incidence of ASD in all 108 assessed children (70 boys, 38 girls) was 6.4% (n = 7). The incidence of ASD in the 52 children with a GID diagnosis was 1.9% (n = 1), which was significantly lower than the incidence of 13% (n = 6) of ASD in the 45 children with a GID-NOS (gender identity disorder not otherwise specified) diagnosis (p<.05). In the 11 children with no GID diagnosis, none had ASD. […]

According to the DISCO algorithms, nine (six boys, three girls) of the 15 ASD suspected adolescents had ASD. The incidence of ASD in all 96 assessed adolescents (45 boys, 51 girls) was 9.4% (n = 9). The incidence of ASD in the 77 adolescents with a GID diagnosis was 6.5% (n = 5), which was significantly lower than the incidence of 37.5% (n = 3) of ASD in the eight adolescents with a GID-NOS diagnosis (p<.05). In the 11 adolescents with no GID diagnosis, one suffered from ASD and transvestic fetishism but no gender dysphoria.

A Bayesian analysis yawns at your puny sample sizes.

Austism prevalence, from Vries [2010].GID-NOS? Sorry, I should unpack those acronyms a bit. “GID” is short for Gender Identity Disorder, what we used to call gender dysphoria until 2010, when clinicians decided it wasn’t an illness. To be diagnosed with it you have to meet a set of criteria. If you didn’t meet those, but clinicians saw some behaviors related to gender that they considered unusual, they could hand out a diagnosis of Gender Identity Disorder Not Otherwise Specified, or GID-NOS to its friends.

Wait. Children and adolescents that showed clear signs of gender dysphoria had a much lower incidence of autism than those with fuzzy signs?

Individuals with an ASD frequently received a GID-NOS diagnosis. GID-NOS appeared to be given when the cross gender behavior and interests were merely subthreshold (mostly in children), or atypical or unrealistic. For example, an adolescent with ASD, who always had the feeling of being different from his peers in childhood, but had no history of childhood cross-gender behavior, became convinced that this feeling of alienation was explained by gender dysphoria. He had the hope that his communication problems would alleviate by taking estrogens.

That kid doesn’t sound dysphoric. They sound like they were told that gender dysphoria and autism are correlated, and on that basis concluded they were dysphoric. At which point a scientific paper used their story as evidence that gender dysphoria is correlated with autism!

By my count, we’re 0 for 4 at showing a correlation between autism and gender dysphoria. The most-cited papers have deep flaws and questionable numbers. Yet because of cheerleaders like Dr. Ken Zucker, and most people’s ignorance of how to read a scientific paper, the claimed correlation has snowballed into a truism and started feeding into itself.

Siobhan’s turn! She’ll discuss Dr. Zucker and his clinic in more detail later this week.