Multivariate Analysis of Insomnia-Related Behaviors Profiled in the Simon's Simplex Collection

Background: Autism spectrum disorder (ASD) is a phenotypically heterogeneous condition with diverse symptomatology and expression of multiple comorbidities. The wide variability in clinical manifestations may be explained by underlying genetic heterogeneity. Incorporating evidence related to expression of comorbidities may identify clinically meaningful subgroups. For example, sleep disorders such as insomnia are common in individuals with ASD, suggesting involvement of shared molecular pathways with sleep or circadian regulation. Identifying subgroups of individuals with co-occurring insomnia may allow detection of clinically relevant genetic mechanisms conferring large effects. 

Objectives: We tested the hypothesis that applying multivariate statistical approaches to analysis of insomnia-related questions available in medical histories from individuals in the Simon’s Simplex Collection would effectively identify subgroups of individuals with ASD expressing comorbid insomnia.

Methods: We conducted cluster analysis evaluating parent responses on five questions specific to insomnia included in the medical histories of 2,708 children ages 4-17 years with ASD. Input variables included parent responses related to whether their children had ever had difficulty going to bed, difficulty falling asleep, frequent or prolonged awakenings at night, sleepwalking or frequent nightmares, and/or needed their parents to lay down with them in order to sleep. We determined differences between the mean sleep duration reported for individuals between identified clusters. We also evaluated differences in the proportion of individuals with parents reporting problems in sleep-related behaviors, and other comorbidities including seizures, attention deficit disorder and constipation. We also assessed whether the proportion of individuals with exonic mutations in two melatonin pathway genes (ASMT and CYP1A2), implicated in risk for both ASD and insomnia, were different between clusters.

Results: Clustering algorithms identified two distinct groups of individuals with ASD. Cluster 1 (the non-insomnia cluster) included 1,898 individuals and Cluster 2 (the insomnia cluster) included 810 individuals. The largest difference in the clusters was related to difficulty falling asleep, followed by whether or not the child needed a parent in order to fall asleep, difficulty going to bed, and frequent or prolonged awakenings. The presence of sleepwalking or nightmares also defined some of the cluster differences; however, differences were substantially less than other questions of interest to insomnia. Individuals in Cluster 2 also had shorter sleep durations than those in Cluster 1. Furthermore, more individuals in Cluster 2 had attention deficit disorder and constipation. There was no difference between the proportions of individuals with mutations in either candidate gene between clusters.

Conclusions: Our clustering algorithm, based on five insomnia-related symptoms assessed in the SSC medical histories, was effective in allowing us to determine the most common sleep concerns in children with ASD.  In addition, presence of insomnia was related to the presence of two other important comorbidities in ASD, ADD and constipation. While we did not observe a difference in mutations in the two candidate genes assessed, it is possible that these clusters represent genetically distinct ASD subsets. In order to fully evaluate potential genetic differences between clusters it will be necessary to evaluate more genes that affect sleep patterns, including circadian and melatonin receptor genes.