Compared to What? Early Brain Overgrowth in Autism and the Perils of Population Norms

Background:  

The presence of a robust association between Autism Spectrum Disorder(ASD) and early brain overgrowth(EBO) is widely accepted, and continues to influence the cutting edge of ASD research. The bulk of published data regarding early brain growth in ASD comes from studies of head circumference(HC), an excellent proxy for brain size in early childhood. 

Objectives:  

To reappraise of the evidence base for EBO in ASD given recent reports that several large independent samples of typically developing children show EBO relative to HC reference norms used by several seminal EBO reports in ASD. 

Methods:  

We systematically review all published HC tests of the EBO hypothesis in ASD (34 studies encompassing ~ 3k ASD and 60k controls participants), and analyze new data from a cohort of 57 preschool-aged male Caucasian children (35 ASD, 22 healthy controls) with ~ 330 longitudinal HC measures between birth and age 18 months.  Throughout, we test if the strength of evidence for EBO depends on the type of control data with which HC data in ASD are compared.  We supplement traditional sources of HC norms in ASD research (e.g CDC) with recently published “Primary Care Norms” (PCNs): the largest (~500k HC measures between birth and 18 months) contemporary set of US-based HC norms.

Results:  

Systematic Review -- Eighty-five percent of all HC studies in ASD make use of HC norms. Cross-sectional studies that do not find evidence for brain enlargement in ASD tend to include a comparison with controls rather than solely relying on HC norms (p=0.0006). Longitudinal HC studies that compare ASD data to CDC norms unanimously identify rapid HC centile increases in ASD between birth and ~age 12 months. In contrast, most longitudinal studies comparing HC growth between ASD participants and controls do not find evidence for EBO in ASD during the first year of life. By integrating prior reports, we robustly confirm the well-replicated pattern of EBO in ASD relative to CDC norms, but show that the timing of this EBO is mimicked by PCN norms.

New Data -- We did not find any cross-sectional differences in raw HC between ASD and controls at any pediatric surveillance time-point in the first 18 months of life, or group differences in raw HC change between time-points. Both ASD and controls groups showed abnormally accelerated HC growth relative to CDC HC norms, but not PCN norms.

Conclusions:  

The best-replicated aspect of EBO in ASD reflect generalizable HC norm biases rather than a disease-specific biomarker. Removing these biases leaves partial support for a subtle divergence of HC growth between a sub-group of clinically-recruited children with ASD and community-recruited controls during the second year life that (i) results in ~5mm group difference in mean HC at 18 months (~ effect size 0.5d), and (ii) may index body size and SES related confounds.  Our findings provide a cautionary reappraisal of the EBO hypothesis in ASD and raise broader concerns about the use of popular growth norms in clinical and academic medicine.