NEW YORK – Combining disease-associated genetic variants with data on early developmental milestones can help predict which children diagnosed with autism spectrum disorder (ASD) will develop intellectual disability (ID), according to new research led by investigators at the University of Montreal and McGill University.
"Our approach builds on previous studies showing that the effects of rare variants are modulated by an individual's genetic background of common variants," senior and co-corresponding author Sébastien Jacquemont, a researcher at the University of Montreal, and his colleagues wrote in a paper published in JAMA Pediatrics on Monday.
In an effort to put together prognostic models for ID in children diagnosed with ASD, researchers from the University of Montreal, McGill, the University of Cambridge, and elsewhere analyzed data from 5,633 children with ASD, who were enrolled through the Simons Foundation Powering Autism Research (SPARK), the Simons Simplex Collection, and MSSNG cohorts, including 4,574 boys and 1,059 girls.
The children had been diagnosed with ASD at an average age of 4 years, the team explained, and had ID assessments between the ages of 8 and 14 years.
"Although early signs of autism are often observed between 18 and 36 months of age, there is considerable uncertainty regarding future development," the authors explained. "Clinicians lack predictive tools to identify those who will later be diagnosed with co-occurring intellectual disability (ID)."
The investigators found that ID prediction was relatively limited using prognostic models based on sets of genetic variants alone. These included collections of common variants in polygenic scores for ASD or variants related to cognitive ability based on prior genome-wide association studies, as well as rare variants of different types with larger individual effects.
Polygenic scores for cognitive ability and ASD were negatively associated with ID, they reported, noting that models based on polygenic scores could be improved with the inclusion of data on de novo missense or loss-of-function variants as well as deletions and duplications implicated in developmental delay.
"We observed that the addition of each class of variant into the model … consistently improved [positive predictive values (PPVs)]," the authors wrote, adding that "integration of all genetic variants also increases the number of identified children among those having ID (i.e., sensitivity) from 6 percent when using only polygenic scores to 30 percent when incorporating all variants, while maintaining a stable PPV of approximately 30 percent."
On the other hand, the investigators found that combining early developmental milestone markers with genetic data offered a boost for identifying ID in children with milestone delays and for ruling out ID more broadly.
"Although the addition of genetic variants to developmental milestones especially improved the identification of individuals who will not develop ID [negative predictive values (NPVs)], the ability to stratify the probabilities of ID using genetic variants was twofold greater in individuals with delayed milestones compared with those with typical development," the authors reported.
After training this model in 4,085 children from the SPARK cohort, the team validated it in another 1,183 SSC participants and 365 participants from the MSSNG study, demonstrating that genetic variants appeared to be particularly informative when combined with early developmental milestones such as walking and first words.
"This prognostic study highlights the feasibility of using predictive models to assist clinicians in their assessment of children referred for autism," the authors suggested. "Models can provide families with probability estimates for different developmental trajectories and the corresponding levels of uncertainty tied to the interpretation of developmental milestones and genetic findings in the context of autism."