Plasma Metabolic Features Associated with Developmental Regression in Autism Spectrum Disorder: A Pilot Untargeted Metabolomics Study

Developmental regression, the loss of previously acquired language and social skills during the second year of life, affects 20–40% of children with autism spectrum disorder (ASD), yet its metabolic underpinnings remain poorly defined. Existing metabolomics studies have focused on amino acid and acylcarnitine alterations, leaving the eicosanoid, sphingolipid, and glycoconjugate axes largely unexamined. Plasma from 25 children with regressive ASD (REG+), 27 with non-regressive ASD (REG−), and 21 typically developing controls (aged 2–6 years) was profiled using untargeted Q-TOF LC/MS (Agilent 6530, ESI+ mode, 100–1700 m/z). From 6657 detected features, a stepwise curation pipeline combining statistical significance (raw p < 0.05 with Benjamini–Hochberg FDR correction, q < 0.05; FC > 1.5), analytical plausibility, endogenous origin verification, and literature-based directional alignment yielded a prioritized set of 14 annotated metabolic features. Exploratory PLS-DA (Partial Least-Squares Discriminant Analysis) visualization showed separation among the three groups within the present dataset; however, diagnostic or predictive performance was not evaluated. The 14-feature panel clustered along four biochemical axes. A feature putatively annotated as prostaglandin E3 (PGE3) showed a 7.13-fold reduction in REG+ versus controls (1.58-fold in REG−; FC REG+/REG− = 4.52). N-Acetylneuraminosyl-(α2-6)lactosamine and aspartylglycosamine were decreased 4.44-fold and 5.42-fold in REG+, respectively, with regression-specific amplification (FC REG+/REG− > 2.0). Six sphingolipid–myelin metabolites showed coordinated biosynthetic depletion and catabolic elevation in both ASD groups without regression specificity. DHEA-S exhibited the largest subgroup differential (FC REG+/REG− = 8.89), with paradoxically greater depletion in the non-regressive group. This exploratory study identified several plasma metabolic features showing abundance differences among children with REG+, REG−, and typically developing controls. These findings are hypothesis-generating and require confirmation using authenticated standards, targeted quantitative methods, and independently recruited cohorts before any diagnostic, predictive, or therapeutic relevance can be established.