The Relationship between Alcohol- and Sleep-related Traits: Results from Polygenic Risk Score Analyses and Mendelian Randomization Studies

Background Epidemiologic studies have shown an association between sleep abnormalities and alcohol-related traits. Recent genome-wide association studies (GWAS) have identified genetic variants associated with sleep-related traits, including insomnia and sleep duration, and with alcohol-related phenotypes, including alcohol use disorder (AUD) and level of alcohol consumption. Objectives We investigated whether genetic risk for insomnia and sleep duration abnormalities are associated with AUD and alcohol consumption. We also evaluated the causal relationships between sleep- and alcohol-related traits. Methods Individual level phenotype and genotype data from the Million Veteran Program was used. Polygenic risk scores (PRS) were computed using summary statistics from two recent discovery GWAS of insomnia (N=453,379 European-ancestry (EA) individuals) and sleep duration (N=446,118 EAs) and tested for association with lifetime AUD diagnosis (cases, N=34,658 EAs) and past-year Alcohol Use Disorders Identification Test-Consumption scale scores (AUDIT-C, N=200,680 EAs). Bi-directional two-sample Mendelian Randomization (MR) analyses assessed causal associations between the two sleep traits and the two alcohol-related traits. Results Insomnia PRS was positively associated with AUD at 2/9 PRS thresholds, with p<0.01 being the most significant (OR = 1.02, p = 3.48 × 10 −5 ). Conversely, insomnia PRS was negatively associated with AUDIT-C at 6/9 PRS thresholds (most significant threshold being p=0.001 (β = - 0.02, p = 5.6 × 10 −8 ). Sleep duration PRS was not associated with AUD, but was positively associated with AUDIT-C at 2/9 PRS thresholds, with the most significant threshold being p = 1 × 10 −6 (β = 0.01, p = 0.0009). MR analyses supported a significant positive causal effect of insomnia on AUD (14 SNPs; beta = 104.14; SE = 16.19; p = 2.22 × 10 −5 ), although with significant heterogeneity. MR analyses also provided nominal evidence of a causal effect of AUD on insomnia (10 SNPs; beta = 0.01; SE = 0.007; p = 0.01). Finally, MR analyses showed that decreased sleep duration had a causal effect on the risk of AUD (27 SNPs; beta = -63.05; SE = 3.54; p = 4.55 × 10 −16 ) and was robust to sensitivity analyses. Conclusion The genetic risk for insomnia shows pleiotropy with AUD, and sleep continuity abnormalities have a causal influence on the development of AUD.