Is autism being overdiagnosed?

NEW YORK – Despite claims by a recent research study that autism is being overdiagnosed, experts at Autism Speaks said that conclusion overstates what the study can realistically draw from its analysis.

The study published last week in JAMA Psychiatry concluded that changes in how providers diagnose autism are blurring the line between people with autism and the general population.

“When looking at any condition, whether we’re talking about autism or any other health condition, we expect that including milder forms of it will decrease the effect size,” said Autism Speaks Chief Science Officer Thomas Frazier. “That is not evidence that the milder forms of the condition are any less part of the condition or that those people don’t still need appropriate care.”

Looking at the “effect size,” or changes in certain variables in research studies over time, this study looked at the cumulative data of many autism studies over about 20 years.

“Effect size is almost always largest in the first published studies reporting a new finding. This is because the initial papers would not be published unless there was a ‘significant’ finding. We have since learned to expect regression to the mean for effect sizes across almost all domains of research,” said Jeremy Veenstra-VanderWeele, M.D., child and adolescent psychiatrist at Columbia University. “We need to be careful to limit our conclusions to the data in front of us.”

Like with any large analysis, the researchers chose which papers to include and which to leave out.

The effect size, such as what the researchers looked at here, depends on how close the measures in the studies are related to the core features of the condition. This study did not include many papers that used measures closely linked to autism — for example, highly specific social communication measures like eye gaze in social and nonsocial context — that might have led to a different effect size.

“While understanding the effect size for autism is a useful way to look at changes in the field over time, this particular study seems to draw broader conclusions than may be warranted because of the types of data the researchers chose to include,” Frazier said.