PHILADELPHIA -- The head of the American Psychiatric Association committee rewriting the diagnostic criteria for autism spectrum disorders took on the panel's critics here, accusing them of bad science.
Susan Swedo, MD, of the National Institute of Mental Health, said a review released earlier this year by Yale University researchers was seriously flawed. That review triggered a wave of headlines indicating that large numbers of autism spectrum patients could lose their diagnoses and hence access to services.
Swedo spoke at the American Psychiatric Association's (APA) annual meeting, in her role as chairperson of the work group developing new diagnostic criteria for neurodevelopmental disorders in DSM-5, the forthcoming fifth edition of the APA's Diagnostic and Statistical Manual of Mental Disorders.
She was especially incensed by reports in consumer media about the Yale group's study, led by a New York Times article with a "blaring" headline that read, "." The Yale study, according to the Times article, found that most patients with Asperger's syndrome and about 25% of those with overt autism would not qualify for those diagnoses under DSM-5.
Bloggers in the autism spectrum community then got the numbers wrong and claimed that DSM-5 would deprive 65% of all autism patients of their diagnoses, "striking fear in the hearts of families," Swedo said.
In fact, she said, the Yale study and hence the Times and other reports messed up. "I can assure you that it is not true," she told APA meeting attendees.
Nevertheless, she said, the panic has now "spun out of control" to the point that a New York state legislator recently introduced a bill that would require physicians in the state to use the current DSM-IV criteria for diagnosing autism.
"That would be a big mistake, because DSM-IV has not been very useful for women, for toddlers, for any child from 3 to 8 years of age, and also was almost useless for adolescents," Swedo said.
She said her group's main goal is actually to make diagnosis easier in these groups, with a second goal of not depriving patients already carrying autism spectrum diagnoses (including Asperger's syndrome) of their eligibility to receive services.
DSM-IV versus DSM-5
Under current criteria for what is called autistic disorder in DSM-IV, delays in social interaction, communication, or normal play activities have to be seen prior to age 3. Consequently, the diagnosis is off-limits for children whose delays weren't noticed until later.
In the draft for DSM-5 -- although Swedo and APA officials emphasized that the document has not been finalized yet -- the strict age requirement is dropped, instead stating that symptoms must have been present in early childhood.
The rationale, according to Swedo, is that although the delays and deficits in autism spectrum disorders are present, probably in infancy and certainly in toddlerhood, they may escape notice until later because young children often aren't required to socialize extensively and parents may provide enough care and support to mask the delays and deficits.
In some cases, children may reach their teens before their social and communication deficits cause them serious trouble.
Other changes to the criteria included a reworked organizational scheme. In DSM-IV, autism disorder, Asperger's syndrome, and two related conditions -- childhood integrative disorder and what was called "pervasive developmental disorder not otherwise specified (PDD-NOS)" -- were separate diagnoses with their own separate criteria.
Swedo's group has proposed to collapse them all into a single "autism spectrum disorder" with one set of checklist criteria coupled with severity ratings in social communication, restricted interests, and repetitive behaviors.
She said a literature review indicated poor specificity and sensitivity for the DSM-IV diagnoses of Asperger's syndrome and PDD-NOS. One study of diagnoses made at 11 different clinics showed that rates of Asperger and PDD-NOS diagnoses varied enormously whereas autism was diagnosed much more consistently.
Another factor arguing in favor of a single autism spectrum disorder, Swedo said, was that the types of symptoms in autism, Asperger's syndrome, PDD-NOS, and childhood integrative disorder were very similar; what differed was the severity or predominance of different symptoms.
The work group also had two motivations for doing away with PDD-NOS as a category. Swedo said the group's view was that "pervasive developmental disorder" is a misnomer insofar as it does not involve deficits or delays in every aspect of development -- in fact, it is largely restricted to social communication.
Moreover, the DSM-5 leadership wanted get rid of "not otherwise specified" categories throughout the manual because such catchall designations conveyed little clinical information. And gaps in the DSM-IV diagnostic scheme had made NOS diagnoses the most popular within many disorder groupings.
The Yale Study
The study by James McPartland, PhD, and colleagues was published last month in the , but drew media attention in January when it was presented at the Icelandic Medical Association, helped along by a Yale press release.
Further spicing the media interest was the fact that its senior author, Fred Volkmar, MD, had been a member of Swedo's work group but quit because he disagreed with the majority's viewpoint.
The group's analysis indicated that 60.6% of patients meeting DSM-IV criteria for an autism spectrum disorder would qualify for one under the DSM-5 draft.
Using data from a 1993 DSM-IV field trial as the reference, the Yale researchers estimated that DSM-5's diagnostic sensitivity was 25% for those with Asperger's syndrome in DSM-IV, 76% for those with autistic disorder, and 28% for those with PDD-NOS.
The group also found that specificity was improved with the DSM-5 criteria. Nevertheless, they warned of "significant public health ramifications regarding service eligibility" if the DSM-5 criteria are adopted.
Swedo said the Yale group misused the 1993 field trial data because it was inappropriate to take clinical evaluations structured a certain way to evaluate the DSM-IV criteria against DSM-III, and use them to determine how the DSM-5 criteria would perform.
"It was not just comparing apples and oranges, it was comparing apples with Apple computers," she argued. "We [in DSM-5] were using words that hadn't really been used in DSM-IV."
If anything, she suggested, the DSM-5 criteria are more sensitive, not less, for picking up the key features of autism spectrum disorder -- particularly "the integration of gestures and verbal communication."
Another point of criticism in the Yale study to which Swedo objected was DSM-5's treatment of social behavior in comparison with DSM-IV.
The relevant DSM-IV checklist item is "failure to develop peer relationships and abnormal social play." In DSM-5, this would be replaced with a dimension scale of "difficulties adjusting behavior to suit different social contexts."
Swedo said this language would not only be sensitive for autism spectrum but would also exclude many children whose primary diagnosis is ADHD, who can also have trouble with peers and social play.
DSM-5 Trials
She pointed to the new field trial data from DSM-5 as justifying the work group's decisions. The criteria showed excellent reliability -- that is, different clinicians evaluating the same child usually came to the same diagnoses, with intraclass kappa values of 0.66 and 0.72 at the two academic centers where the criteria were tested.
Moreover, when the clinicians applied DSM-IV and DSM-5 criteria to the nearly 300 children included in the trial, the autism spectrum prevalence was not changed much.
At Baystate Medical Center in Springfield, Mass., it increased a bit, to 25% with DSM-5 versus 23% with DSM-IV.
At Stanford University in Palo Alto, Calif., it dropped to 20% from 26%. A closer look at the Stanford data, however, suggested that fewer children "lost" a diagnosis than the percentages suggest, according to Swedo.
There were a total of 41 children who qualified for an autism spectrum disorder diagnosis under DSM-IV versus 36 with DSM-5. But there were an additional 10 children who received a diagnosis of a newly proposed condition dubbed "social communication disorder" in the DSM-5 draft, which had no counterpart in DSM-IV.
Among the total of 46 who received diagnoses of autism spectrum or social communication disorders under DSM-5 were six who could not be diagnosed with anything under DSM-IV.
In total, DSM-5 picked up five more children with autism spectrum or social communication disorders than did DSM-IV at the Stanford site, although Swedo cautioned that the small numbers and single site mean the findings can't be generalized.
She emphasized that her work group had determined that its recommended criteria must meet the "first do no harm" test, meaning that people currently receiving services for Asperger's syndrome should continue to be eligible.
Swedo said they sought to create criteria sensitive to include all patients with autism spectrum symptoms, but specific enough to separate those with a genuine disorder -- that is, with clear impairments -- from those such as self-described "Aspies" whose functional abilities appear normal.
Public comment on all draft DSM-5 diagnostic criteria is open until June 15 through the website. The work groups are supposed to submit their final drafts in August. The entire DSM-5 must be finished in December and is slated for formal release in May 2013.