In this study, we used PGS to investigate associations between sleep disturbances and ADHD using a clinical sample of children diagnosed with ADHD and their parents. First, we found that polygenic liability for long sleep duration was over-transmitted from parents to children with ADHD and this finding was supported in a combined analysis using an independent replication sample. Second, we did not find evidence that polygenic liability for sleep disturbances was associated with an increased risk of insomnia, poor sleep quality or hypersomnia in children with ADHD. Finally, we did not find evidence that greater genetic risk for ADHD or other frequently comorbid disorders were associated with increased risk of sleep disturbances in children with ADHD.
We found weak evidence that children with ADHD over-inherit polygenic liability for longer sleep duration. Both long and short sleep duration have previously been phenotypically associated with ADHD . Prolonged sleep duration is a symptom of hypersomnia, which is common in people with ADHD . Other evidence pointing to a shared pathophysiology comes from research on narcolepsy; both narcolepsy and ADHD can be treated using psychostimulants such as methylphenidate  and the prevalence of ADHD in children with narcolepsy is greater than in the general population . Moreover, narcolepsy PGS have been associated with ADHD traits in the general population . One theory is that inattentive and hyperactive-impulsive symptoms are a coping mechanism for deficits in alertness, but research supporting this is mixed . Although our finding of over-transmission of sleep duration polygenic liability was not significant in an independent, larger ADHD sample, the direction of the effect was the same and the result was significant in analysis combining both samples. The effect size was stronger in the primary sample compared to the replication sample or combined analysis. This may be due to differences in the two samples; for example, fewer children in the IMAGE replication sample had sleep problems compared to the primary sample.
We did not find evidence in the primary or replication samples that children with ADHD over- or under-inherit polygenic liability for other sleep traits, specifically insomnia or chronotype. This did not support our hypothesis that sleep disturbances in ADHD arise due to shared genetic influences and contrasts with the previously reported genetic correlation between ADHD diagnosis and insomnia . This could be due to differences in methodology, given we compared polygenic liability for insomnia in children to their parents’ average liability, rather than to screened controls. Our results also contrast with previous PGS studies in children in the general population, which found genetic links between ADHD and sleep phenotypes [9,10,11]. One possibility is that previous results were influenced by the fact that ADHD and sleep problems co-occur in the general population, which was not accounted for. Our results indicate that when potential confounders (e.g. ancestry, socioeconomic background) are taken into account using a trio design, there is limited evidence of shared genetic effects between childhood diagnosed ADHD and sleep phenotypes measured in adults.
Our results also indicated that polygenic liability for sleep disturbances was not associated with an increased risk of insomnia, poor sleep quality, or hypersomnia in children with ADHD. One explanation is that existing GWAS of sleep problems in adults do not index sleep problems in childhood, which extends to sleep problems in the context of clinically-diagnosed ADHD. The GWAS summary statistics we used are the largest available GWAS of sleep phenotypes but were derived from the UK Biobank, which consists of adults aged 40–69 years . There is emerging evidence that the heritability of sleep traits changes across the lifespan [40, 41] and it is possible that the genetic architecture of sleep phenotypes is different in adults compared to children, regardless of ADHD diagnosis. However, currently available sleep GWAS in children are based on much smaller samples (e.g. ). Another possibility is that participants in our primary sample ranged in age from 5 to 18 years, whereas a recent study found that the association between ADHD and sleep disturbances peaked between 8 and 10 years of age . It is, therefore, possible that genetic associations would also be greatest in this age range. In addition, we observed low rates of some sleep disorders in our sample (e.g. only 31 children in our sample had hypersomnia). This could be due to the interview method used to assess sleep phenotypes, or because a high proportion of our sample was taking medication for ADHD or sleep. Rates of hypersomnia might increase as children with ADHD become adults; it would therefore be pertinent to investigate associations between polygenic liability for adult sleep traits and sleep disturbances in adult samples with ADHD.
We did not find support for our hypothesis that greater polygenic liability for ADHD increased the risk of sleep disturbances in children with ADHD. This suggests that unlike other phenotypes (e.g. conduct disorder) , sleep disturbances in children with ADHD do not index a greater burden of risk for ADHD. It is possible that sleep disturbances in ADHD arise due to other factors, which should be investigated further. For example, children with ADHD are more likely to live in families of lower socioeconomic status , which is a risk factor for poor sleep and daytime sleepiness in children . Symptoms of ADHD may make it more difficult for children and their parents to adopt healthy sleep practices, such as regular bedtimes. Randomised controlled trials of sleep interventions that improve sleep practices have been shown to be effective in improving sleep and ADHD symptoms in children with ADHD [6, 7]. Medications to treat ADHD may also improve or disrupt sleep ; however, we observed the same pattern of results in sensitivity analyses adjusting for ADHD medication and excluding children taking medication for sleep.
Other explanations for the link between ADHD and sleep should be considered. First, in our study we examined common genetic variants for sleep traits, but it is possible that sleep disturbances comorbid with ADHD also arise due to rare or de novo genetic variants for sleep traits or ADHD. Second is the role of gene-environment interplay; sleep disturbances in children with ADHD could arise in those who have a genetic predisposition to sleep problems and are exposed to particular environmental risk factors. Third is the role of power as the PGS were based on examining a small fraction of the genome (tens of thousands of SNPs) and our sample is much smaller than those typically used in GWAS and genome-wide genetic correlation analyses. Finally, disorders such as ASD, oppositional defiant disorder, anxiety, and depression are highly comorbid with ADHD [46, 47]. These disorders are also associated with sleep problems [48, 49] and may drive associations between ADHD and sleep disturbances. However, in this study, we did not find evidence that sleep problems in children with ADHD were driven by polygenic liability for ASD, MDD or anxiety. Future research needs to explore the impact of these comorbidities on sleep disturbances in children with ADHD.
Our study has several limitations, in addition to those outlined above. We measured sleep phenotypes using an interview, and a combination of parent and child-ratings. Some studies have shown discrepancies between parent and child reports of sleep, and their association with genetic factors; Breitenstein et al.  found that the heritability of sleep phenotypes in children differed depending on measurement method (self-report, parent-report or actigraphy). Future research should explore associations between ADHD and sleep using multiple measures of sleep phenotypes.
Our study also lacked phenotypic measures of chronotype, sleep duration, and other sleep disorders previously associated with ADHD , therefore we could not examine associations between ADHD/sleep PGS and these phenotypes. We were unable to test in our sample whether children with ADHD (regardless of sleep difficulties) had a higher burden of polygenic liability for sleep phenotypes, as we did not have a comparison sample of children without ADHD. Sleep disturbances were assessed when 30% of the sample were taking medication for ADHD and 15% were taking sleep medication, which may have influenced our results, although our sensitivity analyses indicated this is unlikely to fully explain our results. Our study may also have been underpowered due to only having 328 trios in the primary sample and 844 trios in the replication sample. Detailed sleep phenotypes were not available in the replication sample.
Sleep problems are common in children with ADHD and impact on illness course, quality of life and family relationships [2,3,4]. Understanding why these sleep problems occur is important for understanding the aetiology of ADHD and identifying individuals who may be at higher risk of developing these problems. Using two large samples of children with ADHD, we find limited evidence for over-transmission of polygenic liability for sleep phenotypes, with weak evidence that polygenic liability for long sleep duration may be over-transmitted from parents to children. Our results also suggest comorbid sleep problems are not driven by genetic factors linked to sleep phenotypes measured in adults, or driven by polygenic liability for ADHD or other commonly comorbid neurodevelopmental and psychiatric disorders. However, our analyses utilised data from adult sleep GWAS and need to be extended using data from sleep GWAS in children, once sufficiently large samples are available. Further exploration of the role of sleep comorbidities in individuals with ADHD, and examination of this relationship using a wider range of sleep phenotypes is also needed. Our findings also highlight the need for additional exploration of biological and environmental factors that may give rise to sleep disturbances in children with ADHD.