Gray matter covariations in autism: out-of-sample replication using the ENIGMA autism cohort.

Only people who already had an IQ score were included, which the authors say likely skews the sample. IQ and sex were not matched between the autistic and non-autistic groups, or between the two datasets, even though the analysis adjusted for both. The 34 sites used different scanners, field strengths and scan settings, and collected their samples differently. This is the authors' own leading explanation for why the result did not replicate. Little clinical information was available in the ENIGMA dataset: co-occurring conditions such as ADHD and anxiety and medication use were largely unknown, and there was too little continuous behavioural data to relate brain patterns to behaviour. The study measures brain structure only. It reports no function, symptoms or treatment response, so it cannot say anything about what helps autistic people.

The work was funded by public and charitable sources: the China Scholarship Council, the European Union's Horizon 2020 and Seventh Framework programmes, the Netherlands Organization for Scientific Research (a VICI grant), and the Wellcome Trust. No commercial sponsor or treatment manufacturer is named.

Do not read this as proof that autistic people have no brain differences. The study failed to reproduce two specific patterns, and the group difference disappeared once variation between scanning sites was taken into account, so the non-replication says as much about pooling sites as about autism itself.