Atypical structural connectivity of language networks in autism spectrum disorder: A meta-analysis of diffusion tensor imaging studies.

Most pooled results showed large differences between studies, and there were too few studies of the ventral tracts or of adults to work out what was driving that. The included studies used very different scanners, sequences and analysis methods, and the authors could not separate the effect of those differences from the findings. The participants were mostly male and mostly high-functioning: 31 of the 33 studies included only people with IQ above 70, and 13 studied males only. The findings may not hold for people with intellectual disability, or for women and girls. More than 80% of the studies came from the United States or Europe, so other populations are barely represented. Children under 3 were excluded, and no study followed the same people over time, so nothing here shows how these tracts change with age in an individual. The review could not test whether the scan differences relate to actual language performance, because the original studies measured language in too many different ways to combine. DTI cannot resolve fibres that cross each other, so the measures are an imperfect picture of the tracts.

The authors declare no conflicts of interest. The work re-analyses previously published data, so it needed no ethics approval.

Do not read the child-versus-adult difference as these brain differences repairing themselves with age. The children and the adults came from different studies scanned at a single time point, not the same people followed over years, so the age comparison is between groups of studies rather than a developmental trajectory. Nor does this show that weaker white matter connections cause language difficulty in autism; nothing here was designed to test cause, and no treatment was studied.