Review of: Rødgaard EM et al, JAMA Psychiatry 2019;76(11):1124–1132

Study Type: Meta-analysis of meta-analyses

We’ve heard a lot about the increasing prevalence of autism, up from 1 in 2,000 in 1965 to 1 in 45 in 2015. What we have not heard are definitive answers as to why. This study used a novel design to explore whether the research has included an increasing number of mild cases along the autism spectrum in recent decades.

The authors examined differences between subjects identified with autism vs healthy controls across neurocognitive measures. The magnitude of the difference was measured by effect size, and they assessed how those effect sizes changed in studies dating back to the 1980s. For comparison, the authors conducted the same analysis in studies of schizophrenia, a condition whose prevalence (1%) has not changed over time. The autism data were based on 11 meta-analyses involving 27,723 subjects and 7 neurocognitive constructs. For schizophrenia, they used 4 meta-analyses involving 10,159 subjects and 3 neurocognitive constructs.

For the autism group, the differences between the controls and those diagnosed with autism shrank over time in all 7 neurocognitive constructs. Those differences reached statistical significance for 5 of those constructs: emotion recognition, theory of mind, executive planning, the event-related potential P3b on EEG, and brain size. Between 2000 and 2015, the effect sizes for these 5 measures decreased by 45%–80%. In contrast, no changes in effect sizes were found when comparing neurocognitive constructs related to schizophrenia.

The changes in the autism studies correlated with the year of the study and were not explained by changes in study design or by inclusion of high-functioning autism in the samples. Although DSM-5 introduced new criteria for autism, most of the changes observed in this study occurred during the DSM-IV era (1994–2013).

TCPR’s Take
It looks like our concept of autism has broadened in the past 20 years. Rates of diagnosis are on the rise, and research studies are including people with autism who look increasingly like healthy controls. When a patient presents with this diagnosis, look carefully at how it was made. Milder cases will likely benefit from a non-pathologizing approach that emphasizes their strengths and weaknesses.

To learn more, listen to our 7/27/20 podcast, “How to Talk to Patients With Autism.” Search for “Carlat” on your podcast store.