Psychiatric symptom patterns in autism do not map neatly onto the general population model
In adolescents with autism, psychiatric symptoms do not arrange themselves in the same tidy structure seen in non-autistic youth. This preprint compared Child Behavior Checklist profiles and genetic associations in 11,111 autistic children with verbal abilities from SPARK and 11,041 non-autistic children from ABCD, all between 8 and 15 years old. The investigators were asking a basic but clinically consequential question: whether the emotional and behavioral symptoms that travel with autism reflect the same underlying architecture seen in the broader pediatric population, or whether autism reshapes that architecture at both the phenotypic and genetic level.
The answer was not subtle.
Confirmatory factor analysis fit the canonical CBCL structure in ABCD, but not in SPARK. In practical terms, the usual population-wide way of organizing internalizing, externalizing, attention, and related symptom domains described the non-autistic cohort well, yet it failed to capture autistic adolescents with the same fidelity. That disconnect matters at the bedside because assessment tools built on general-population assumptions may blur clinically distinct symptom clusters in autism, especially when social behavior, attention, developmental delay, and somatic complaints overlap in ways that are not separable by standard subscales alone.
The CBCL holds in non-autistic youth, but autism rearranges the symptom map
The CBCL has long been used as a broad measure of child psychiatric symptomatology, but this analysis shows that its canonical structure cannot simply be carried over into autistic adolescents without adjustment. In ABCD, the factor structure remained intact. In SPARK, it fractured. The most striking departure was not a total loss of order; rather, the broad internalizing domain still emerged, yet other features redistributed across factors in ways that suggest autism-specific symptom organization.
One especially important finding was the emergence of a distinct Developmental Delays factor in SPARK. Social problems also dispersed across multiple factors instead of clustering cleanly into a single dimension. That pattern fits the clinical reality of autism, where social withdrawal, communication differences, cognitive profile, adaptive function, and psychiatric distress can intertwine without collapsing into a single conventional syndrome. A child who appears socially avoidant may be expressing anxiety, cognitive burden, autistic social communication differences, or some combination of all three, and the factor structure here suggests that the usual psychiatric taxonomy may not disentangle those layers well.
Genetic loading diverged between autistic and non-autistic cohorts
Genetics followed the same general direction of divergence. Psychiatric polygenic score associations with canonical CBCL subscales were weaker in SPARK than in ABCD. The clearest example involved attention problems: the ADHD polygenic score association with Attention problems was substantially lower in SPARK (β=0.05 [CI: 0.04, 0.06]) than in ABCD (β=0.17 [0.13, 0.21]). For clinicians and researchers, that difference suggests that the relationship between common inherited risk and observable attention symptoms is altered in autism, rather than simply amplified or reduced uniformly across the board.
This is where the biology becomes especially informative. If a symptom cluster is driven more strongly by common psychiatric liability, it may behave more like the same symptom in the general population. If it is driven less strongly by that liability and more by autism-linked developmental biology, then the phenotype becomes less interchangeable with conventional psychiatric categories. The data point toward exactly that split.
Developmental delay tracked with de novo burden, not common psychiatric polygenic risk
Exploratory factor analysis within SPARK replicated the canonical internalizing factors, but it also exposed additional structure. The Developmental Delays factor showed the strongest associations with de novo burden in autism-associated genes and with lower cognitive ability. At the same time, it was only weakly associated with psychiatric polygenic scores. This combination is clinically meaningful because it implies that some parts of the psychiatric presentation in autism are anchored more in rare, high-impact neurodevelopmental variation than in the common polygenic signals that often inform psychiatric genetics in the general population.
A one-size-fits-all model of psychopathology misses that distinction. Developmental delay in an autistic adolescent is not merely a backdrop variable; it may be part of the same etiologic pathway that shapes symptom expression, making it less likely to line up with conventional genetic predictors built from broader psychiatric cohorts.
The same analysis also showed that common genetic variation explained larger proportions of variance in somatic and inattentive factors and had stronger polygenic score associations there. In other words, some symptom dimensions in autism still appear to be partly governed by the broader inherited architecture of psychiatric traits, while others are more tightly linked to autism-specific developmental genetics. The result is a mosaic rather than a monolith.
Why autism-specific assessment may need to replace borrowed psychiatric assumptions
The clinical implications reach beyond statistical fit. Psychiatric screening in autism often relies on tools that were validated mainly in general pediatric populations. If the underlying factor structure differs, then the same score may not mean the same thing in an autistic adolescent as it does in a non-autistic one. A high attention-problems score, for example, may reflect a different mix of attentional control, sensory overload, developmental delay, or social-communication strain than it would in a non-autistic peer.
That has consequences for diagnosis, treatment selection, and the interpretation of research studies that pool autistic and non-autistic participants without testing whether the symptom architecture itself is equivalent. A factor that appears clinically broad in one group may conceal several distinct pathways in another. In autism, those pathways may be split between common psychiatric risk, rare developmental burden, and cognitive ability in ways that standard checklists were never designed to parse.
What this preprint can and cannot establish
The study carries important strengths: a large sample, direct comparison between autistic and non-autistic youth, and parallel evaluation of phenotypic structure and genetic correlates. It also has boundaries that matter. The autism cohort included children with verbal abilities, so the findings may not extend to the full spectrum. The work is a preprint, so the results have not yet undergone peer review. And while the analysis suggests autism-specific psychiatric architecture, it does not by itself provide a ready-made diagnostic replacement for existing tools.
Even so, the core message is hard to ignore. Psychiatric symptoms in autism are not simply the general-population symptom model with more noise added. The structure differs, the genetic correlates differ, and the clinical interpretation should differ with them.