Resting-state fMRI reveals hyper- versus hypo-connectivity autism subtypes in 400-participant cohort
Stanford fMRI study of 400 autistic youth identifies two connectivity-based subtypes with distinct clinical profiles. Cross-sectional design limits causal inference; subtype-stratified trials needed next. Evidence supports moving beyond broad autism categories toward biologically grounded subtypes.
The team applied unsupervised clustering to connectivity matrices derived from 10-minute eyes-open scans, revealing a high-connectivity subtype with elevated thalamocortical coupling and a low-connectivity subtype showing reduced long-range frontoparietal links. Both groups met DSM-5 autism criteria yet differed significantly on ADOS social scores and sensory processing measures. Sample size provided 80% power to detect medium effect sizes in connectivity differences after multiple-comparison correction.
This pattern aligns with prior smaller studies reporting connectivity extremes but adds rigorous subtype validation through cross-validation and replication in an independent 150-participant hold-out set. It challenges unitary models of autism as a single connectivity disorder and suggests targeted interventions: excitatory modulation for the hypo-connected group and inhibitory approaches for the hyper-connected group.
Key limitation is the cross-sectional design at a single site; longitudinal data are required to test subtype stability and treatment response. Multi-site RCTs with pre-specified subtype stratification would substantially strengthen causal claims for precision therapies.
Stanford team: Within 36 months, at least one phase-2 trial will report subtype-specific response rates differing by >25 percentage points at p<0.01.
Sources (2)
- [1]Primary Source(https://www.nature.com/articles/s41586-024-08234-5)
- [2]Supporting Source(https://www.cell.com/neuron/fulltext/S0896-6273(23)00412-8)