Linking White-Matter Development to Clinical Variation in Autism: Longitudinal Normative Modelling of FA in the EU-AIMS LEAP Cohort.

Background: Autism is characterised by marked neurobiological and clinical heterogeneity, which often limits the sensitivity of traditional group comparisons. Longitudinal normative modelling offers a way to capture heterogeneity and to describe an individual's brain-behaviour relationships beyond diagnostic labels. Methods: We analysed diffusion MRI scans from 3 waves of the EU-AIMS Longitudinal European Autism Project (LEAP). Using normative models of white matter fractional anisotropy (FA) we computed individual deviations relative to age typical norms. Our sample included 544 different individuals across all waves (Wave 1: 344 [autistic 189]; Wave 2: 297 [autistic 162]; Wave 3: 226 [autistic 145]), aged 7-37 years, with mean intervals of ∼1.5 years (W1-W2) and ∼6 years (W1-W3). We first tested for group level differences between autism and no-autism and then examined the associations with sensory, adaptive, cognitive, and behavioural measures across development within the autism group. Results: Although mean FA differences between autistic and non-autistic participants were not significant, correlations between individual differences in FA and behaviour across repeated measurements yielded moderate but reproducible results. The cingulum bundle, uncinate fasciculus, callosal fibres (genu and splenium), and the superior longitudinal fasciculus were among the most frequently identified pathways associated with clinical traits. For these circuits, FA values were higher in individuals with better adaptive functioning (VABS) and lower in those reporting sensory processing (SSP) difficulties. Findings were mixed for associations between FA and symptom severity (ADOS). The largest cross-sectional signal appeared at Wave 2, and although some of the same circuitry re-appeared over about 7.5 years from Wave 1 to Wave 3, effect sizes remained modest and variable, consistent with heterogeneity and power differences across waves, but still pointing to a stable white matter tract pattern. Conclusions: Autism's white-matter organisation reflects a dimensional and individually patterned architecture rather than a categorical distinction. Stable, person-specific deviations from shared neurodevelopmental pathways are linked to behavioural variation among autistic individuals. Longitudinal normative modelling therefore provides a practical approach to understand and quantify this heterogeneity.