Multimodality Integration of Neural Social Activation and Social and Language Scores Reveals Three Replicable Profound and Milder Autism Subtypes With Divergent Clinical Outcomes

Abstract

ABSTRACTSocial affective and communication symptoms stand at the center of autism, and usually become apparent within the first 1-3 years of life. Symptom severity differs widely across toddlers and clinical outcomes, ranging from near-neurotypical to poor. The biological bases of this early and wide symptom diversity are largely unknown. While more than two dozen studies have attempted to subgroup early-age clinical heterogeneity, most studies fail to rigorously validate discovered subtypes using multiple methods, and none linked observed clinical subtypes with underlying functional neural signatures. Using a well-established approach for precision medicine patient subtyping (Similarity Network Fusion) and multiple rigorous validation methods, we integrated thoroughly replicated measures of social neurofunctional activation and social and language ability in 137 toddlers at early ages. Results identified three distinct social neural-clinical ASD subtypes, validated using multiple methods. One subtype was consistent with a ‘profound’ autism profile with negligible social neural activation, severe social and language symptoms, low social interest, and little clinical improvement. Another ASD subtype had a contrasting pattern with only mildly reduced social neural activity, near neurotypical social and language abilities, and substantial age-related clinical improvement. One principal implication of these results is that the “spectrum” of ASD heterogeneity is not truly a continuous spectrum from the neurobiological and clinical perspective. The profound autism subtype is the neurofunctional, clinical and developmental opposite of the mild ASD subtype, suggesting different etiological mechanisms. A second implication is that neurobiological and clinical subtype differences highlight the need to develop subtype-specific treatments, particularly for the profound subtype. Third, treatment studies with an undetermined mix of subtypes could fail or succeed based on how many patients from each subtype are included in the mix.