Integrating causal discovery and clinically-relevant insights to explore directional relationships between autistic features, sex at birth, and cognitive abilities

Angela Tseng, Sunday M. Francis, Eric Rawls, Christine A. Conelea, Nicola M. Grissom, Erich Kummerfeld, Sisi Ma, Suma JacobPsychological 2025 · DOI

Abstract

Our investigation unveiled sex-specific directional relationships that inform our understanding of differing needs and outcomes associated with biological sex in autism and may serve to further development of targeted interventions.

Introduction

Autism Spectrum Disorder (ASD) is characterized by significant heterogeneity, yet research has historically been skewed by a 4:1 male-to-female prevalence ratio. This "male-centric" focus has left the unique presentations of autistic females—who often employ "camouflaging" strategies—poorly understood. Beyond simple correlations, identifying the directional mechanisms or "flow" between symptoms like repetitive behaviors, social communication, and cognitive ability is essential for developing effective clinical interventions. This study utilizes Causal Discovery Analysis (CDA) to map these directional pathways and determine how they differ by biological sex.

Methodology

The researchers employed a sophisticated data-driven pipeline to analyze the Simons Simplex Collection (SSC) v15.3, a dataset comprising 2,509 individuals (2,175 males and 334 females). To ensure a fair comparison despite the sample size imbalance and IQ differences, a "Matched Male" (MM) group (N=331) was created, matched to the female cohort by age and IQ.

The analysis involved four key steps:

  1. Exploratory Factor Analysis (EFA): Condensing 161 behavioral items into 15 distinct factors (e.g., "Motor overflow," "Socioemotional unresponsiveness").
  2. Greedy Fast Causal Inference (GFCI): A causal discovery algorithm used to search for the best-fitting directed acyclic graph (DAG) and detect latent common causes.
  3. Structural Equation Modeling (SEM): Estimating the effect sizes of the identified causal relationships.
  4. Bootstrap Resampling: Running the analysis 1,000 times to ensure the stability and consistency of the discovered edges.

Results

Common Causal Pathways

Several directional relationships were found to be universal across all groups. Non-verbal IQ (NVIQ) emerged as a primary "causal ancestor," where higher NVIQ directly led to a reduction in repetitive hand/body movements and sensory preoccupations. Additionally, increasing age was strongly associated with a decrease in "Motor overflow" (excessive impulsive activity), with an effect size of approximately -0.39 for females.

Causal model illustrating the directional flow of symptoms common to both females and all males.

Sex-Specific Drivers

A key finding was the role of Verbal IQ (VIQ) as a sex-specific driver. In females, higher VIQ was linked to lower levels of repetitive speech and higher socioemotional awareness. This suggests that verbal ability may be a critical tool used by females to navigate social structures and potentially facilitate "masking" or camouflaging. This relationship was not significantly present in the male groups.

Behavioral Cascades and Intervention Targets

The analysis identified a stable behavioral chain: Motor Overflow → Oppositional Behaviors → Self-Injurious Behaviors. This pathway was consistent across females and matched males, identifying "motor overflow" as a high-value upstream target for early intervention to mitigate the risk of later self-harm.

Specific causal paths comparing females (F) and matched males (MM), highlighting the stability of behavioral chains.

Furthermore, the model indicated that as autistic females age, they may experience an increase in socioemotional unresponsiveness. This likely reflects the rising complexity of social demands during adolescence, which may eventually outpace a girl's ability to compensate or camouflage.

Discussion and Limitations

While CDA provides powerful insights into directionality, this study is limited by its cross-sectional nature; longitudinal data would be required to confirm true temporal causality. Additionally, the use of the "Simplex" dataset (families with only one autistic child) and reliance on parent-report questionnaires may introduce specific biases. Future research should apply these causal models to longitudinal cohorts to track how these symptom networks evolve from childhood into adulthood.

Summary updated

Abstract

Our investigation unveiled sex-specific directional relationships that inform our understanding of differing needs and outcomes associated with biological sex in autism and may serve to further development of targeted interventions.
BibTeX
@inproceedings{tseng2025integrating,
  author = {Angela Tseng and Sunday M. Francis and Eric Rawls and Christine A. Conelea and Nicola M. Grissom and Erich Kummerfeld and Sisi Ma and Suma Jacob},
  title = {{Integrating causal discovery and clinically-relevant insights to explore directional relationships between autistic features, sex at birth, and cognitive abilities}},
  booktitle = {Psychological},
  year = {2025},
  doi = {10.1017/s0033291725000571},
}
Figure