Transdiagnostic Neurocognitive Difficulties Map onto Neuroanatomical Signatures and Predict ADHD Trajectories in Early Development: a clustering study

Background Autism frequently co-occurs with attention-deficit/hyperactivity disorder (ADHD), which significantly impacts developmental outcomes and quality of life. While executive functions (EF) and reward processing differences are commonly observed in both conditions, it remains unclear whether early neurocognitive variation can predict emergence of ADHD traits in preschool children. This study aimed to identify transdiagnostic neurocognitive profiles in a preschool sample and evaluate their clinical, neurobiological, and prognostic relevance. Method This study utilised data from a well-characterised cohort of 240 autistic and neurotypical preschoolers aged 34–52 months. Indices from a novel touchscreen battery assessing inhibitory control, sustained attention, and reward learning were entered into a density-based clustering algorithm. The resulting subgroups were externally characterised on concurrent clinical and neuroanatomical characteristics, and as predictors of ADHD features after 12–15 months. Results The clustering analysis identified four neurocognitive subgroups: EF+Reward difficulties, showing challenges across all assessed domains alongside elevated autistic and ADHD traits but typical-range IQ; Low Completion, marked by task completion difficulties, lower IQ, and moderate ADHD traits; and two subgroups with relatively intact profiles, exhibiting typical neurocognitive performance and low ADHD traits. The EF+Reward difficulties and Low Completion subgroups showed widespread reductions in frontal and parietal cortical surface area relative to the intact groups. Subgroup membership also predicted ADHD traits over time, supporting the prognostic value of these neurocognitive profiles. Limitations: External replication was not possible due to the absence of comparable publicly available preschool cohorts. Multiple internal and external validation strategies nonetheless support subgroup robustness. Additional limitations include reliance on parent-reported ADHD measures and a small EF+Reward difficulties subgroup. Conclusions Neurocognitive subtypes, derived from scalable innovative tools, can parse heterogeneity in early development in autistic and neurotypical sample and predict early ADHD features, mapping onto relevant neuroanatomy. Our study highlights the potential of early transdiagnostic cognitive profiling for identifying children at an increased likelihood of co-occurring autistic and ADHD traits and informing targeted early interventions.