This is the internet weblog of Bix Frankonis.

The Autism ‘Recovery’ Study That Wasn’t?

Last month, a study dropped (Google cache because U.K. Ed Chat has since deleted their article) in Journal of Child Neurology brazenly entitled, “When an Early Diagnosis of Autism Spectrum Disorder Resolves, What Remains?”

“It’s certainly encouraging to confirm that a subset of children with early ASD diagnosis accompanied by developmental delays can in essence recover from the disorder and go on to have typical social and cognitive functioning,” said lead author Lisa Shulman, M.D., professor of pediatrics at Einstein and interim director of the Rose F. Kennedy Children’s Evaluation and Rehabilitation Center (CERC) at Montefiore. “But by and large, these children continue to struggle with daily life. Almost all of them still have to contend with language and learning disabilities and a variety of emotional and behavioral problems.”

The short version: the records of 569 autistic kids were studied, from diagnosis around two years old to followup around six years old. The “vast majorty” of them had received some form of early intervention, such as Applied Behavior Analysis. At study’s end, only 7% (38 children) were deemed to have “recovered” while nonetheless having to “contend with language and learning disabilities and a variety of emotional and behavioral problems”; only 3 kids (about 0.5% of the original 569) were deemed to have “recovered” without having to contend with other problems.

According to the study’s lead author, Lisa Sherman, this shows that it is possible to cover from autism.

What I don’t get is how that’s the inevitable conclusion, rather than the conclusion to be drawn only if you first make some initial–and very circular–assumptions. To wit: you first have to assume that recovery is possible in order for these results to tell you that recovery is possible.

There’s another way to read these results: early intervention treatments, such as ABA, don’t work. For one thing, more than 90% of the kids apparently continued to be diagnosed as autistic at the end of the study period, calling into qustion the effectiveness of ABA-style interventions.

More interestingly, perhaps: the fact that 7% of the kids “recovered” but nonetheless had to contend with myriad other problems doesn’t immediately mean that recovery is possible. Those numbers instead also could be looked at to mean, worst case scenario, that early intervention treatments such as ABA actually damage autistic kids. (There’s only correlation here, but it’s at least as valid an explanation based upon these results as is “recovery”.)

There are some revealing bits in the Wall Street Journal article published shortly after the study was released.

She says it’s hard to determine if these children who outgrow their diagnoses do so because of their own natural progression or in response to early intervention. The children who do usually have the mildest cases. Another possible explanation: Some children were misdiagnosed.

Children who are no longer diagnosed with ASD show improvements with communicating and interacting with other people, which can include more eye contact and an easier time with back-and-forth conversations.

Unpack this and you see the problem making some initial assumptions can cause.

Let’s take the last part first, which assumes that children who become capable of mimicking neurotypical communication behavior must necesssarily have “recovered” from autism rather than, perhaps, merely been forced to suppress the ways in which their atypical neurology normally and naturally would be expressed.

Then let’s back up to the previous paragraph. Shulman assumes that “outgrowing” the diagnosis only can be a result of “natural progression” or “early intervention” or an original “misdiagnosis”. It can’t possibly, to her mind, be the result of the forceful suppresion of the kids’ natural neurological differences while those differences themselves still remain.

It’s an echo of how Shulman restricts the potential options to be considered in the study paper itself.

Though it has been reported in multiple studies that a small subset of children with early ASD improve in terms of their social functioning and no longer warrant a diagnosis of autism, this persistent finding continues to beg several questions: Was autism overcalled in these children to begin with? Are some children better able to respond to intervention? Does the specific intervention the child receives contribute to outcome? All are possible. Based on our experience, our sense is that the symptoms evolve—in some children in response to intervention, in others due to their individual developmental trajectories.

Rather, if what’s happening through some form of early interventions is that we are forcing a person with one neurology to outwardly behave as if they have a different neurology, Shulman’s results very easily can be used to draw the conclusion not that recovery is possible but that this sort of forced artificial, superficial recovery is itself contributing to leaving some kids with “language or learning disabilities”, “externalizing behaviour problems”, “internalising behaviour problems”, or “a significant mental health diagnosis”.

The missing data here, of course, is the early intervention treatment breakdown across those 569 kids, including the 7% of kids who “recovered” with problems, and the 3 kids who “recovered” without problems. We don’t know, specifically, what sorts of treatment each segment received. That data isn’t just missing here, it’s missing in the published paper itself.

Which means Shulman’s assumptions have blocked even the rest of us from examining how those different early interventions could feed into the results.

You can’t engage in a study like this assuming that recovery is possible. You can accept that many parents and caregivers believe it is, and then study what’s happening when they try, but you can’t begin with the assumption that recovery is possible and then only look at the results through that lens. You have to examine all the possible reasons for the results you get.

One last point, also from the WSJ piece.

Children who are later diagnosed with disorders like psychosis or anxiety may have had them all along. But such conditions are hard to diagnose before children can speak well, Dr. >Shulman notes.

This is one of the glaring deficiencies of studies like these: the subjects can’t speak for themselves, to speak to what these early intervention treatments have or have not done for them or to them. It’s almost always based on the external evaluations not just of doctors but or caregivers–which is one of the reasons “behaves like a typical kid” is deemed to be a success for early interventions.

I hope other researchers are looking skeptically at studies like this, whose results do not inherently yield what their authors claim.