{"paper_id":"3f809cbf-fc2c-4d1f-8d65-c8416e3e4683","body_text":"Identifying the potential role of insomnia on \nmultimorbidity: A Mendelian randomization phenome-wide \nassociation study in UK Biobank  \nMark J Gibson1 2, Deborah A Lawlor1 3*, Louise AC Millard1 3* \n* these authors contributed equally \n \n* Corresponding author: mark.gibson@bristol.ac.uk \n \n1 MRC Integrative Epidemiology Unit (IEU) at the University of Bristol, Bristol, United \nKingdom \n2 School of Psychological Science, University of Bristol, Bristol, United Kingdom \n3 Department of Population Health Sciences, Bristol Medical School, University of \nBristol, Bristol, United Kingdom \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n \n \nAbstract \nObjectives To identify the breadth of potential causal effects of insomnia on health \noutcomes and hence its possible role in multimorbidity. \nDesign Mendelian randomisation (MR) Phenome-wide association study (MR-\nPheWAS) with two-sample Mendelian randomisation follow-up. \nSetting Individual data from UK Biobank and summary data from a number of \ngenome-wide association studies. \nParticipants 336,975 unrelated white-British UK Biobank participants. \nExposures Standardised genetic risk of insomnia for the MR-PheWAS and \ngenetically predicted insomnia for the two-sample MR follow-up, with insomnia \ninstrumented by a genetic risk score (GRS) created from 129 single-nucleotide \npolymorphisms (SNPs). \nMain outcomes measures 11,409 outcomes from UK Biobank extracted and \nprocessed by an automated pipeline (PHESANT). Potential causal effects (i.e., those \npassing a Bonferroni-corrected significance threshold) were followed up with two-\nsample MR in MR-Base, where possible. \nResults 437 potential causal effects of insomnia were observed for a number of \ntraits, including anxiety, stress, depression, mania, addiction, pain, body \ncomposition, immune, respiratory, endocrine, dental, musculoskeletal, \ncardiovascular and reproductive traits, as well as socioeconomic and behavioural \ntraits. We were able to undertake two-sample MR for 71 of these 437 and found \nevidence of causal effects (with directionally concordant effect estimates across all \nanalyses) for 25 of these. These included, for example, risk of anxiety disorders \n(OR=1.55 [95% confidence interval (CI): 1.30, 1.86] per category increase in \ninsomnia), diseases of the oesophagus/stomach/duodenum (OR=1.32 [95% CI: \n1.14, 1.53]) and spondylosis (OR=1.57 [95% CI: 1.22, 2.01]).  \nConclusion Insomnia potentially causes a wide range of adverse health outcomes \nand behaviours. This has implications for developing interventions to prevent and \ntreat a number of diseases in order to reduce multimorbidity and associated \npolypharmacy.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \n \nIntroduction \nWhile there is still much debate over the exact purpose of sleep, it is clear that sleep \nis vital for healthy functioning and likely to be multifaceted. Experiments on rats have \nsuggested that sleep is linked to antioxidative enzyme levels in the brain which \nregulate the levels of reactive oxygen species (by-products of the metabolization of \noxygen which damage cells) (1). It has also been proposed that sleep is vital for the \nconsolidation of information, learning and memory (2, 3). Insomnia is defined as \nregular dissatisfaction with the quality or quantity of sleep for a prolonged period and \nincludes difficulty initiating or maintaining sleep (4). Evidence suggests that 6-7% of \nthe European population have a diagnosis of insomnia, while 33-37% self-report \nhaving insomnia symptoms (5-7). It is the second most prevalent mental health \nWhat this paper adds \nWhat is already known on this topic \n• Insomnia symptoms are widespread in the population and insomnia is the second \nmost common mental health disorder after anxiety disorders. \n• Insomnia might have broad and numerous effects on health and multimorbidity, but \nboth observational and Mendelian randomisation studies have focused on select \nhypothesised associations/effects rather than taking a systematic hypothesis-free \napproach across many health outcomes.  \nWhat this study adds \n• This study uses a hypothesis-free approach to systematically identify causal effects \nof insomnia on 11,409 health outcomes; it identified 437 potential causal effects \nand for 71 that could be followed-up with two-sample MR, 25 showed evidence of a \ncausal effect and were directionally consistent across all analyses. \n• These findings identify potential routes for developing interventions to prevent and \ntreat a number of diseases and reduce multimorbidity and polypharmacy.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \ndisorder (after anxiety disorder) and is more common in women and the elderly (6, \n7). Multimorbidity, defined as patients living with two or more chronic health \nconditions, is associated with polypharmacy, poor quality of life and premature \nmortality (8, 9). It is increasingly recognised as a threat to global health and \nidentifying potential causes of multimorbidity is a research priority (10). Given the \nhigh prevalence of insomnia symptoms, and their association with many diseases \n(including depression (11), diabetes (12), hypertension (13), dementia (14) and \ncardiovascular disease (15, 16)), insomnia could be a cause of multimorbidity. \nHowever, associations with disease outcomes may not be due to a causal effect of \ninsomnia on the outcome and could reflect residual confounding or reverse causality \nfrom undiagnosed prevalent disease (17). Furthermore, studies to date have focused \non hypothesised selected outcomes, predominantly mental, neurocognitive and \ncardiometabolic outcomes, rather than systematically, using a hypothesis free \napproach, searching for potential causal effects across a wide range of health and \ndisease outcomes. \nMendelian randomisation (MR) is a method used for testing causal relationships, that \ngenerally uses genetic variants that are robustly associated with the exposure of \ninterest as instrumental variables (IV) (18). MR is typically less prone to confounding \nof the exposure outcome association and reverse causation than conventional \nobservational epidemiology; as genetic variation is determined at conception it \ncannot be altered by disease status (19). However, it has other potential sources of \nbias, in particular those due to weak instruments, confounding of the instrument-\noutcome association and horizontal pleiotropy (20) (the core assumptions of MR \nhave been previously reported in detail (21)). Previous MR studies have provided \nsome evidence that insomnia may lead to heavier substance use (22, 23), increased \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nBMI, an increased risk of type 2 diabetes (24), cardiovascular diseases (25), autism \nspectrum disorder, bipolar disorder (26), pain (27), and major depressive disorder \n(28), and also increased levels of an inflammatory marker (glycoprotein acetyls) and \ncitrate, but with no strong evidence that it causes widespread metabolic disruption \nacross 115 metabolic markers (29). However, insomnia may have a causal effect on \nmany health outcomes beyond those already studied. If insomnia is a cause of \nmultimorbidity then insomnia treatments, such as the UK national institute for health \nand care excellence (NICE) guideline (30) recommended cognitive behavioural \ntherapy-Insomnia (31), would prevent a range of other adverse health outcomes. \nMR-PheWAS combines both MR and Phenome Wide Association Studies \n(PheWAS) (32) to explore causal relationships with many phenotypes in a \nhypothesis-free manner (33). To our knowledge only one previous study has \nundertaken an MR-PheWAS of insomnia (34). In that study the automated tool \nPhenoScanner (35) was used to explore causal effects of insomnia on 179 outcomes \n(34). It identified 478 potential causal effects (using on a p-value threshold of 5 × 10\n-\n8) including on adiposity, mental health, musculoskeletal, respiratory/allergic and \nreproductive phenotypes. However, that MR-PheWAS was part of an illustrative \nexample in a methodological paper fousing on addressing one of the MR \nassumptions, and none of the potential causal effects were explored further with \nreplication or sensitivity analyses. Here we use an open-source software package \ncalled PHESANT (36) to conduct a large-scale MR-PheWAS in UK Biobank (37, 38), \nto search for novel causal effects of insomnia on health outcomes. The authors \nfollowed the STROBE-MR reporting guidelines when writing this paper (39) and this \nstudy was not pre-registered.   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nMethods \nStudy Population \nWe used data from UK Biobank, a large prospective cohort study (dataset ID 43017 \nof UK Biobank application 16729, phenotypic data extracted on 24/02/2021). UK \nBiobank recruited 503,325 adults aged between 37–73 years. They were recruited \nbetween 2006 and 2007 and attended one of 22 test centres across the UK. Of the \n503,325 participants, genetic data (40) was successfully obtained for 487,406 \nparticipants (Supplementary Text). Participants were then excluded from this sample \nif they did not meet the genetic quality control (41), they were not of white-British \nancestry, they were not part of the maximal subset of individuals not related to any \nother individual to the third degree or higher, or they had since withdrawn their \nconsent (as of 09/08/2021). The remaining 336,975 participants were included in the \nMR-PheWAS (See Figure 1 for a flow diagram). \nGenetic Risk Score \nWe generated a weighted genetic risk score (GRS) using 129 independent single-\nnucleotide polymorphisms (SNPs) previously identified to associate with insomnia \n(Supplementary text) at GWAS significance (with p< 5 × 10\n-8) in 23andMe, Inc. (24). \nThese data were requested from 23andMe as they were not provided in the original \nGWAS paper. SNPs were weighted by their per-allele association with insomnia in \nthe original GWAS (Supplementary Table S1). We used a linkage disequilibrium (LD) \nthreshold of R2>0.001 to clump the GWAS significant SNPs into independent SNPs. \nLD was calculated in the 1000 Genomes European data (42) and the TwoSampleMR \n(MR-base) R package v0.5.6 (43) was used to clump GWAS significant SNPs into \nindependent SNPs. One SNP (rs28458909) was not available in UK Biobank and \nthus was replaced by a proxy (rs28780988) that was in close LD (R2 = 1). All \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \npalindromic SNPs had an effect allele frequency (EAF) falling below 0.49 or above \n0.51 in UK Biobank and 23andMe and therefore could be harmonised. \nOutcomes \n11,409 outcome variables were derived and analysed using PHESANT (36). \nOutcomes included those obtained from responses to baseline and follow-up \nquestionnaires, baseline assessments such as weight, height, blood pressure and \nDXA scan bone density measurements, follow-up assessments such as \naccelerometer measurements and a range of different scans (including brain and \ncardiac scans), biomarker measures from blood or urine samples and outcomes \nfrom linkage to primary and secondary care, and the national cancer and death \nregisters. In order to summarise our overall findings from the MR-PheWAS, \noutcomes were assigned to categories (e.g., Mental Health) based on their UK \nBiobank category (e.g., Online follow-up > Mental health > Anxiety). Measurements \nthat were not health related outcomes were assigned to the Auxiliary Variables \ncategory. These included outcomes such as hospital administration records and \nprocedural metrics such as the length of time taken to complete the touch screen \nquestionnaire. Individual sleep variables from the mental health and physical health \ncategories were then reassigned to a sleep category and medication variables in the \nphysical health category that were for mental disorders were reassigned to the \nmental health category. We then manually assigned outcomes in these two \ncategories to subcategories. Supplementary Table S3 shows which category and \nsubcategory each UK Biobank category is assigned to. \nPheWAS Analysis \nThe PHESANT v1.0 package was used for the MR-PheWAS. We adjusted for age at \nassessment, sex and the top 10 genetic principal components to control for \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \npopulations stratification (44). A complete case analysis was undertaken by \nPHESANT meaning participant numbers differ between outcomes and we chose to \nexclude outcomes with less than 100 cases. PHESANT derives outcomes from the \nUK Biobank data and defines whether they are continuous, binary, ordered \ncategorical or unordered categorical and tests the association with a trait of interest, \nin our case the insomnia GRS, using linear (using inverse normal rank transformed \ndata to ensure a normal distribution), logistic, ordered logistic, and multinomial \nlogistic regression, respectively. The results are presented as difference in mean SD \nof inverse rank normal transformed continuous outcomes and odds ratio (OR) for \ncategorical outcomes, per 1 standard deviation (SD) increase in the weighted GRS. \nWe defined potential causal effects as any insomnia GRS-outcome association that \npassed the Bonferroni-corrected significance threshold of 4.38x10-6 (0.05/11,409) in \nthe MR-PheWAS. The less conservative false discovery rate (FDR) correction was \nalso calculated and reported but was not used to identify potential causal effects for \nfollow-up. \nSensitivity Analysis \nAs the SNPs used to construct the GRS in the main analysis are not replicated there \nis a higher chance that spurious SNPs could have been falsely detected, increasing \nthe chance that the GRS may contain horizontally pleiotropic SNPs. We created two \nsensitivity analysis GRS which used SNPs which were replicated in UK Biobank, \nmeaning the presence of spurious SNPs is less likely. However, as the MR-PheWAS \nwas conducted in UK Biobank this overlap between the selection and test sample \ncould introduce bias through overfitting or winner’s curse. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nThese GRSs included 111 SNPs (Supplementary Table S4) identified in a meta-\nanalysis GWAS of both 23andMe and UK Biobank (24). The first sensitivity GRS \n(S1) weighted SNPs by the per-allele association with insomnia from the pooled UK \nBiobank and 23andMe analyses. The second sensitivity GRS (S2) weighted by the \n23andMe per allele associations with insomnia. While 114 independent SNPs were \nidentified in the original GWAS 3 SNPs were removed for the following reasons. One \nSNP (rs9540729) was palindromic and had an EAF in 23andMe between 0.49 and \n0.51 meaning it could not be aligned with the UK Biobank data. Therefore, it was \nexcluded from both scores for consistency. Two SNPs, rs77641763 and \nrs117630493, had point estimates in the meta-analysis GWAS which were in the \nopposite direction to the 23andMe GWAS, and so were also removed. Of the SNPs \nincluded in the score, 38 were also used in the main GRS. The authors of the \noriginal GWAS used UK Biobank data to calculate LD and an LD threshold of \nR2>0.001 was used to define independent SNPs.  \nFollow-up two-sample MR \nWe undertook follow-up analyses using two-sample MR for all outcomes for which \nthe association with the GRS was identified as a potential causal effect of insomnia. \nThe purpose of this was to confirm the reliability of the potential causal effects \nidentified in the MR-PheWAS and to provide a causal estimate. The TwoSampleMR \n(MR-base) v0.5.6 (43) was used to conduct the follow-up. It was decided a priori that \noutcomes included in the auxiliary variables or sleep categories would not be \nfollowed-up. First, we conducted an automated search for relevant GWAS using pre-\nspecified search terms for each outcome. The search automatically excluded GWAS \nthat included solely UK Biobank data, included non-European populations or \nstratified by sex, based on the meta-data included in the MR-Base database. Of the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nremaining GWAS, we excluded those that did not match a follow-up outcome on \nmanual inspection, those for which the origins of the data used could not be \ndetermined, and those that used UK Biobank or 23andMe data. If the only GWAS \navailable for a particular outcome included UK Biobank or 23andMe data (but did not \nonly include UK Biobank or 23andMe data) we undertook follow-up in those GWAS \nand report the extent of overlap between the two samples. Of the remaining GWAS \nwe then chose the most suitable for a given trait, that was either of the most suitable \nmatch in terms of the trait used or where multiple GWAS had suitable traits we chose \nthe one with the larger sample size. \nThe two-sample MR analysis used the 129 SNPs used to construct the GRS in the \nmain analysis. The SNP-insomnia associations were extracted from the \n23andMe/UK Biobank meta-analysis GWAS summary data (24), and the SNP-\noutcome associations were extracted from the GWAS for each outcome. The \nTwoSampleMR (MR-base) package attempted to harmonise SNPs and excluded \nthose it could not (e.g. if a suitable proxy cannot be found for missing SNPs or if \nSNPs were palindromic with allele frequencies near to 0.5). We used the inverse-\nvariance weighted (IVW) method for our main two-sample MR analyses (45), and \nweighted median regression MR (46) and MR-Egger (47) as sensitivity analyses to \nexplore potential bias due to unbalanced horizontal pleiotropy. Weighted median MR \nis unbiased when less than 50% of the weight is made of horizontally pleiotropic \nSNPs, while MR-Egger is unbiased when the magnitude of horizontal pleiotropy of \nthe included SNPs is not proportional to the effect of each SNP on the exposure. All \ncode can be found at https://github.com/MRCIEU/PHESANT-MR-PheWAS-Insomnia \nv1.0.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nResults \nMR-PheWAS \nThe insomnia GRS was associated with an increased risk of insomnia in UK \nBiobank: Odds Ratio (OR) of report of usually versus never/rarely/sometimes having \ntrouble falling or staying asleep = 1.08 [95% Confidence Interval (CI): 1.07, 1.09] per \none standard deviation higher GRS (p=3.59x10\n-84, McFadden’s pseudo R2=0.01). \nSee Supplementary Figure S1 for the association of each SNP with insomnia. \nOf the 11,409 associations included in the MR-PheWAS, 437 were identified as \npotential causal effects (Supplementary Table S3). These included anxiety, stress, \ndepression, mania, addiction, pain, body composition, immune, respiratory, \nendocrine, dental, musculoskeletal, cardiovascular and reproductive traits, as well as \nsocioeconomic and behavioural traits. Figure 2 shows the proportion of potential \ncausal effects of insomnia by broad categories of outcomes. For associations \nbetween insomnia and mental health outcomes 96 of 301 (32%) were identified as \npotential causal effects. There were higher proportions of these in 10 out of 17 of the \nmental health subcategories (Figure 3), including depression (38%), anxiety (48%), \ngeneral (33%), well-being (87%), suicide and self-harm (24%), and mania (19%). \nExamples of adverse potential causal effects on mental health outcomes are having \nseen a doctor for nerves, anxiety, tension or depression (OR=1.06 per 1SD higher \nweighted GRS [95% CI: 1.05, 1.07]), neuroticism (OR=1.05, [95% CI: 1.04, 1.05]), \nmood (OR=1.05, [95% CI: 1.04, 1.05]) and the frequency of depressed moods in the \nlast two weeks (OR=1.05, [95% CI: 1.04, 1.06]). By contrast none of the outcomes in \nsubcategories of psychosis, personality disorders, organic brain disorders, eating \ndisorders, developmental disorders or behavioural syndromes had associations with \nthe GRS which were identified as potential causal effects.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nOf the physical health category 197 out of 6451 (3%) associations with the insomnia \nGRS were identified as potential causal effects. Higher proportions of potential \ncausal effects (Figure 4) were seen for the pain (30%) and body composition (19%) \nsubcategories. Examples of adverse potential causal effects on physical health \noutcomes were long-standing illness, disability or infirmity (OR=1.07, [95% CI: 1.06, \n1.07]), back pain in the last month (OR=1.04, [95% CI: 1.04, 1.05]), body fat \npercentage (mean difference=0.01, [95% CI: 0.01, 0.02]), and gastro-oesophageal \nreflux (OR=1.05, [95% CI: 1.04, 1.06]).  \nFor the family and childhood category 17 out of 96 (18%) associations were \nidentified as potential causal effects. This category included some outcomes that \ncould not be plausibly affected by adult insomnia, and might reflect shared family \n(inherited) predisposition to insomnia and its potential causal effects on fertility and \nhealth outcomes across family members, such as maternal smoking around birth \n(OR=1.03, [95% CI: 1.02, 1.04]), number of full brothers (OR=1.02 [95% CI: 1.02, \n1.03]), mother's age at the time the participant was recruited (mean difference=-0.01, \n[95% CI: -0.02, 0.01]) and severe depression in a sibling (OR=1.06, [95% CI: 1.04, \n1.07]).  \nExamples of outcomes identified as potential causal effects from the \nlifestyle/behaviours category (44 out of 854, 5%) are age first had sexual intercourse \n(mean difference=-0.02, [95% CI: -0.03, -0.02]), time spent watching television \n(OR=1.04, [95% CI: 1.03, 1.05]) and pack years of smoking (mean difference=0.03, \n[95% CI: 0.02, 0.03]). Examples of outcomes identified as potential causal effects \nfrom the sociodemographic category (38 out of 1053, 4%) are average total \nhousehold income before tax (OR=0.96, [95% CI: 0.96, 0.97]) and age completed full \ntime education (OR=0.96, [95% CI: 0.95, 0.97]). There were 2 of 2160 (0.1%) \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \noutcomes identified as potential causal effects from the brain imaging category. \nAlternatively, the brain/cognition category had no potential causal effects. Full details \nof the numbers in each category/subcategory and the numbers and percentages of \noutcomes in those categories that are potentially influenced by insomnia are \nprovided in Supplementary Tables S4 and S5.\n \nSensitivity Analysis \nThe two GRS used in sensitivity analyses were also associated with an increased \nrisk of insomnia in UK Biobank: OR=1.11 [95%CI: 1.10, 1.12] per one standard \ndeviation increase in the S1 GRS (p=1.65x10\n-157, McFadden’s pseudo R2=0.01); \nOR=1.10 [95% CI: 1.10, 1.11] per one standard deviation increase in the S2 GRS \n(p=6.76x10-145, McFadden’s pseudo R2=0.01). The correlations between each score \nwere strong (R2=0.99 [95% CI: 0.99, 0.99] for the S1 GRS with the S2 GRS; R2=0.69 \n[95% CI: 0.69, 0.69] for the S1 GRS with the main GRS; R2=0.70 [95% CI: 0.69, \n0.70] for the S2 GRS with the main GRS). See Supplementary Figure S2 for \nassociations between each SNP and insomnia for the S1 and S2 GRS. \nFor GRS S1 and S2, 498 and 490 associations were identified as potential causal \neffects respectively. There was considerable concordance between the three GRS, \nwith 72% of the 542 associations identified using at least one GRS identified using \nall three. For associations identified as potential causal effects in any of the three \nGRS, the association was directionally consistent across all three (Supplementary \nFigure S3 and Supplementary Table S3). \nFollow-up two-sample MR \nOf the 437 potential causal effects identified in the MR-PheWAS, we identified 71 \nwith a relevant GWAS in MR-Base, and hence eligible for follow-up (see Figure 5 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nand Supplementary Tables S6 – S8). Of these, 36 outcomes showed evidence of an \neffect of being a self-reported insomnia case versus not (see Supplementary Text for \n23andMe definition) in the IVW MR analyses, having 95% confidence intervals which \nexcluded the null (Figures 6 and 7 and Supplementary Tables S9 and S10). These \nestimates were in the same direction as the MR-PheWAS for 34 of these 36, with 25 \n(10 continuous and 15 binary) of these having effect estimates in the same direction \nacross all 2-sample MR analyses although with confidence intervals often including \nthe null. These 25 outcomes include a range of categories: mental health outcomes \nsuch as anxiety and post-traumatic stress disorder; body composition outcomes \nsuch as body fat percentage and waist circumference; musculoskeletal outcomes \nsuch as spondylosis, dorsalgia and unspecified arthrosis; the digestive health \noutcomes diaphragmatic hernia and diseases of the oesophagus, stomach and \nduodenum; the respiratory outcomes asthma and bronchitis; and outcomes which \nwere not related to others in the set such as C-reactive protein levels, migraine and \nhigh-density lipoprotein (HDL) cholesterol. Cochran’s Q showed evidence of \nbetween SNP heterogeneity (p<0.05) in both the IVW and MR-Egger analyses for \nnine of these 25 outcomes: asthma, body fat percentage, body mass index, hip \ncircumference, waist circumference, C-reactive protein level, diaphragmatic hernia, \nmigraine and HDL cholesterol, but none showed evidence of horizontal pleiotropy in \nthe MR-egger intercept (see Supplementary Tables S9 and S10 for the Cochran Q \nand MR-egger results for all outcomes).  \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nDiscussion \nIn this study we conducted an MR-PheWAS of insomnia using a GRS of 129 \ninsomnia-associated SNPs and 11,409 outcome variables, using a subsample of \n336,975 unrelated white-British participants from UK Biobank. Of these GRS-\noutcome associations, 437 met our criteria for being potential causal effects, of \nwhich 71 were possible to follow-up using two-sample MR. Follow-up analyses \nshowed consistent evidence of an adverse causal effect of insomnia on 25 outcomes \nincluding those related to anxiety disorders, respiratory disorders, musculoskeletal \ndisorders, disorders of the digestive system and body composition measurements. \nSeveral of the two-sample MR results were underpowered (sample sizes ranged \nbetween 1,000 and 360,838 for outcome GWAS, see Supplementary Table S7) and \nwith larger sample sizes some may be identified as having precisely estimated \ncausal effects of insomnia symptoms. This includes outcomes for which the IVW \npoint estimate was clinically important, such as colon cancer and bipolar disorder. \nTogether with the potential causal effects that we were not able to follow-up, these \nfindings support a role for insomnia in multimorbidity, and the possibility that effective \ninsomnia treatments, such as the cognitive behavioural therapy-Insomnia (31), would \nprevent a range of other adverse health outcomes. Further work is needed to \nunderstand the biological pathways through which the genetic variants may act, and \ntherefore which results are subject to horizontal pleiotropy, as well the mechanisms \nwhich might lead to potential causal effects of insomnia on certain health outcomes. \nFurther research is also needed to establish whether the bidirectional causal effects \nexist between insomnia and the outcomes identified in this work. \nOur results indicate that insomnia has numerous and broad effects on health and \nreplicate previous MR studies (in different samples) which indicate insomnia has \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nadverse effects on neuroticism (48), HbA1C levels (increasing) (49), joint and back \npain/disorders (27), body composition (24), migraine (50) and alcohol use (51). The \nfinding that lifetime predisposition to insomnia increased C-reactive protein levels (a \nbiomarker for inflammation), aligns with previous experimental research which found \na consistent effect of sleep deprivation (52). We also found evidence of novel \npotential adverse causal effects of insomnia on anxiety disorders, allergic disease \n(asthma, hay fever or eczema), asthma, bronchitis, soft-tissue disorders, shoulder \nlesions, diseases of oesophagus/stomach/duodenum, oesophagitis, \ngastroesophageal reflux disease and diaphragmatic hernia. Evidence was also found \nfor an adverse effect of insomnia on post-traumatic stress disorder, where previous \nMR studies found no clear evidence (53), and an increasing effect on HDL \ncholesterol, in the opposite direction to previous MR findings (29).  \nStrengths and Limitations \nA key strength of our hypothesis-free MR-PheWAS is that it allows for many potential \nnovel causal effects of insomnia to be identified. As multimorbidity is increasingly \nidentified in young and older people, research into how it, and associated \npolypharmacy, may be prevented, is becoming increasingly important. MR-PheWAS \nhas the potential to identify exposures that could be targets to prevent multimorbidity \n(8-10). Furthermore, we used two-sample MR to follow-up as many of the potential \ncausal effects as possible and included sensitivity analyses to explore potential bias \ndue to horizontal pleiotropy. \nWe used a Bonferroni corrected p-value threshold to avoid identifying many potential \ncausal effects that are chance findings, but this may be overly stringent and mean \nthat several outcomes for which insomnia has a true but small causal effect may not \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nhave met this threshold due to a lack of statistical power (which differs between \noutcomes). Also, 366 (84%) potential causal effects could not be followed-up \nbecause we were unable to identify a GWAS with summary GWAS data available \nusing MR-Base. Hence, novel potential causal effects still need to be confirmed \nthrough further research.  \nUK biobank has measured a large number of characteristics and has extensive \nlinkage to health records. It also has a large sample size which helps to offset the \nmultiple testing burden of a MR-PheWAS. However, the response rate for UK \nBiobank was 5.5% and those recruited were on average healthier with lower levels of \nchronic diseases than the UK population as a whole, which may have resulted in \nselection bias influencing the MR-PheWAS results (54). Finally, to avoid bias due to \npopulation stratification our analyses were restricted to White-Europeans and we \ncannot assume that our findings would generalise to other ancestral groups. \nConclusion \nOur results suggest that insomnia may have broad effects on health. In particular, we \nidentified novel effects (that replicated in follow-up analyses) on anxiety disorders, \nallergic disease, respiratory disorders, soft-tissue disorders and digestive disorders, \nand confirmed previously identified effects on mental health, hyperglycaemia, pain \nand body composition outcomes. These findings support a role for insomnia in \nmultimorbidity, and the possibility that effective insomnia treatments would prevent a \nrange of other adverse health outcomes. With more and larger GWAS it might be \npossible to replicate other potential causal effects that we were unable to replicate or \nobtain precise estimates for.   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nReferences \n1. Ramanathan L, Gulyani S, Nienhuis R, Siegel JM. Sleep deprivation decreases \nsuperoxide dismutase activity in rat hippocampus and brainstem. Neuroreport. \n2002;13(11):1387-90. \n2. Morin A, Carrier J, Dostie V, Doyon J. Differences in time- and sleep-dependent \nlearning and memory consolidation of motor sequence and visuo-motor adaptation skills. \nSleep. 2005;28:A355-A. \n3. Stickgold R. Sleep-dependent memory consolidation. Nature. 2005;437(7063):1272-\n8. \n4. American Psychiatric Association. Diagnostic and statistical manual of mental \ndisorders: DSM-5. 5th ed. ed. Arlington, VA, Washington, D.C.2013. \n5. Wittchen HU, Jacobi F, Rehm J, Gustavsson A, Svensson M, Jonsson B, et al. The \nsize and burden of mental disorders and other disorders of the brain in Europe 2010. Eur \nNeuropsychopharm. 2011;21(9):655-79. \n6. Morphy H, Dunn KM, Lewis M, Boardman HF, Croft PR. Epidemiology of insomnia: a \nlongitudinal study in a UK population. Sleep. 2007;30(3):274-80. \n7. Ohayon MM. Epidemiology of insomnia: what we know and what we still need to \nlearn. Sleep Medicine Reviews. 2002;6(2):97-111. \n8. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of \nmultimorbidity and implications for health care, research, and medical education: a cross-\nsectional study. Lancet. 2012;380(9836):37-43. \n9. Jani BD, Hanlon P, Nicholl BI, McQueenie R, Gallacher KI, Lee D, et al. Relationship \nbetween multimorbidity, demographic factors and mortality: findings from the UK Biobank \ncohort. BMC Med. 2019;17. \n10. Whitty CJM, MacEwen C, Goddard A, Alderson D, Marshall M, Calderwood C, et al. \nRising to the challenge of multimorbidity. Bmj-Brit Med J. 2020;368. \n11. Baglioni C, Battagliese G, Feige B, Sp iegelhalder K, Nissen C, Voderholzer U, et al. \nInsomnia as a predictor of depression: A meta-analytic evaluation of longitudinal \nepidemiological studies. J Affect Disorders. 2011;135(1-3):10-9. \n12. Cappuccio FP, D'Elia L, Strazzlillo P, Miller MA. Quantity and Quality of Sleep and \nIncidence of Type 2 Diabetes - A systematic review and meta-analysis. Diabetes Care. \n2010;33(2):414-20. \n13. Meng L, Zheng Y, Hui RT. The relationship of sleep duration and insomnia to risk of \nhypertension incidence: a meta-analysis of prospective cohort studies. Hypertens Res. \n2013;36(11):985-95. \n14. de Almondes KM, Costa MV, Malloy-Diniz LF, Diniz BS. Insomnia and risk of \ndementia in older adults: Systematic review and meta-analysis. J Psychiatr Res. \n2016;77:109-15. \n15. Li M, Zhang XW, Hou WS, Tang ZY. Insomnia and risk of cardiovascular disease: A \nmeta-analysis of cohort studies. Int J Cardiol. 2014;176(3):1044-7. \n16. Sofi F, Cesari F, Casini A, Macchi C, Abbate R, Gensini GF. Insomnia and risk of \ncardiovascular disease: a meta-analysis. Eur Heart J. 2012;33:947-. \n17. Fewell Z, Davey Smith G, Sterne JA. The impact of residual and unmeasured \nconfounding in epidemiologic studies: a simulation study. Am J Epidemiol. 2007;166(6):646-\n55. \n18. Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology \ncontribute to understanding environmental determinants of disease? Int J Epidemiol. \n2003;32(1):1-22. \n19. Davey Smith G, Lawlor DA, Harbord R, Timpson N, Day I, Ebrahim S. Clustered \nenvironments and randomized genes: A fundamental distinction between conventional and \ngenetic epidemiology. Plos Med. 2007;4(12):1985-92. \n20. Davey Smith G, Ebrahim S. Mendelian randomization: prospects, potentials, and \nlimitations. Int J Epidemiol. 2004;33(1):30-42. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \n21. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: \na guide, glossary, and checklist for clinicians. Bmj-Brit Med J. 2018;362. \n22. Pasman JA, Smit DJA, Kingma L, Vink JM, Treur JL, Verweij KJH. Causal \nrelationships between substance use and insomnia. Drug Alcohol Depend. 2020;216. \n23. Gibson M, Munafo MR, Taylor AE, Treur JL. Evidence for Genetic Correlations and \nBidirectional, Causal Effects Between Smoking and Sleep Behaviors. Nicotine Tob Res. \n2019;21(6):731-8. \n24. Jansen PR, Watanabe K, Stringer S, Skene N, Bryois J, Hammerschlag AR, et al. \nGenome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and \nfunctional pathways. Nat Genet. 2019;51(3):394-403. \n25. Larsson SC, Markus HS. Genetic Liability to Insomnia and Cardiovascular Disease \nRisk. Circulation. 2019;140(9):796-8. \n26. Gao X, Meng LX, Ma KL, Liang J, Wang H, Gao Q, et al. The bidirectional causal \nrelationships of insomnia with five major psychiatric disorders: A Mendelian randomization \nstudy. Eur Psychiat. 2019;60:79-85. \n27. Broberg M, Karjalainen, J., FinnGen, O llila, H. M. Mendelian randomization highlights \ninsomnia as a risk factor for pain diagnoses. Sleep. 2021;zsab025. \n28. Cai L, Bao YR, Fu XQ, Cao HB, Baranova A, Zhang XR, et al. Causal links between \nmajor depressive disorder and insomnia: A Mendelian randomisation study. Gene. 2021;768. \n29. Bos MM, Goulding NJ, Lee MA, Hofman A, Bot M, Pool R, et al. Investigating the \nrelationships between unfavourable habitual sleep and metabolomic traits: evidence from \nmulti-cohort multivariable regression and Mendelian randomization analyses. BMC Med. \n2021;19(1). \n30. The National Institute for Health and Care Excellence. Insomnia [Internet]. London: \nNICE;  [updated 2021 March; cited 2021 December 6]. Available from: \nhttps://cks.nice.org.uk/topics/insomnia/\n. \n31. Trauer JM, Qian MY, Doyle JS, Rajaratnam SMW, Cunnington D. Cognitive \nBehavioral Therapy for Chronic Insomnia A Systematic Review and Meta-analysis. Annals of \nInternal Medicine. 2015;163(3):191-204. \n32. Denny JC, Ritchie MD, Basford MA, Pulley JM, Bastarache L, Brown-Gentry K, et al. \nPheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene-disease \nassociations. Bioinformatics. 2010;26(9):1205-10. \n33. Millard LAC, Davies NM, Timpson NJ, Tilling K, Flach PA, Smith GD. MR-PheWAS: \nhypothesis prioritization among potential causal effects of body mass index on many \noutcomes, using Mendelian randomization. Sci Rep-Uk. 2015;5. \n34. Yang Q, Sanderson E, Tilling K, Borges MC, Lawlor DA. Exploring and mitigating \npotential bias when genetic instrumental variables are associated with multiple non-exposure \ntraits in Mendelian randomization. medRxiv. 2019;19009605. \n35. Staley JR, Blackshaw J, Kamat MA, Ellis S, Young R, Butterworth AS. \nPhenoScanner: A Database of Human Genotype-Phenotype Associations. Genet Epidemiol. \n2016;40(7):664-. \n36. Millard LAC, Davies NM, Gaunt TR, Smith GD, Tilling K. Software Application Profile: \nPHESANT: a tool for performing automated phenome scans in UK Biobank. Int J Epidemiol. \n2018;47(1):29-35. \n37. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK Biobank: An \nOpen Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of \nMiddle and Old Age. Plos Med. 2015;12(3). \n38. Thompson SG, Willeit P. UK Biobank comes of age. Lancet. 2015;386(9993):509-10. \n39. Skrivankova VW, Richmond RC, Woolf BAR, Davies NM, Swanson SA, \nVanderWeele TJ, et al. Strengthening the reporting of observational studies in epidemiology \nusing mendelian randomisation (STROBE-MR): explanation and elaboration. BMJ. \n2019;375(n2233). \n40. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank \nresource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-9. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \n41. Mitchell RE, Hemani G, Dudding T, Corbin L, Harrison S, Paternoster L. UK Biobank \nGenetic Data: MRC-IEU Quality Control, version 2, 18/01/2019. 2019. \n42. Altshuler DM, Durbin RM, Abecasis GR, Bentley DR, Chakravarti A, Clark AG, et al. \nA global reference for human genetic variation. Nature. 2015;526(7571):68-74. \n43. Hemani G, Zhengn J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-\nBase platform supports systematic causal inference across the human phenome. eLife. \n2018;7. \n44. Mitchell RE, Elsworth B, Mitchell R, Raistr ick C, Paternoster L, Hemani G, et al. MRC \nIEU UK Biobank GWAS pipeline version 2. 2019. \n45. Burgess S, Butterworth A, Thompson SG. Mendelian Randomization Analysis With \nMultiple Genetic Variants Using Summarized Data. Genet Epidemiol. 2013;37(7):658-65. \n46. Bowden J, Smith GD, Haycock PC, Burgess S. Consistent Estimation in Mendelian \nRandomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet \nEpidemiol. 2016;40(4):304-14. \n47. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid \ninstruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. \n2015;44(2):512-25. \n48. Zhang FQ, Baranova A, Zhou C, Cao HB, Chen J, Zhang XR, et al. Causal \ninfluences of neuroticism on mental health and cardiovascular disease. Hum Genet. 2021. \n49. Barry C, Liu JX, Richmond R, Rutter MK, Lawlor DA, Dudbridge F, et al. Exploiting \ncollider bias to apply two-sample summary data Mendelian randomization methods to one-\nsample individual level data. Plos Genet. 2021;17(8). \n50. Daghlas I, Vgontzas A, Guo Y, Chasman DI, Saxena R. Habitual sleep disturbances \nand migraine: a Mendelian randomization study. Ann Clin Transl Neurol. 2020. \n51. Pasman JA, Smit DJA, Kingma L, Vink JM, Treur JL, Verweij KJH. Causal \nrelationships between substance use and insomnia. Drug Alcohol Depend. 2020;214. \n52. Meier-Ewert HK, Ridker PM, Rifai N, Regan MM, Price NJ, Dinges DF, et al. Effect of \nsleep loss on C-reactive protein, an inflammatory marker of cardiovascular risk. J Am Coll \nCardiol. 2004;43(4):678-83. \n53. Lind MJ, Brick LA, Gehrman PR, Duncan LE, Gelaye B, Maihofer AX, et al. Molecular \ngenetic overlap between posttraumatic stress disorder and sleep phenotypes. Sleep. \n2020;43(4). \n54. Munafo MR, Tilling K, Taylor AE, Evans DM, Smith GD. Collider scope: when \nselection bias can substantially influence observed associations. Int J Epidemiol. \n2018;47(1):226-35. \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFootnotes \nAcknowledgements: This research was conducted using the UK Biobank resource \nunder Application Number 16729. This research also used data supplied by \n23andMe under a confidentiality agreement. We would like to thank the research \nparticipants and employees of 23andMe, Inc. for making this work possible \nFunding: This work was supported by a Medical Research Council (MRC) PhD \nstudentship to MJG (grant code: MC_UU_00011/7) and the British Heart Foundation \n(AA/18/7/34219). DAL is further supported by a British Heart Foundation Chair \n(CH/F/20/90003) and National Institute of Health Research Senior Investigator award \n(NF-0616-10102). LACM is supported by a University of Bristol Vice-Chancellor’s \nfellowship. All three authors work in a Unit that is funded by the University of Bristol \nand Medical Research Council (MC_UU_00011/1, MC_UU_00011/6 and \nMC_UU_00011/7). \nThe funders had no role in the study design, collection or analysis of data, or \ninterpretation of results. The views expressed in this paper are those of the authors \nand not necessarily any funder or acknowledged person/institution. \nContributor and guarantor information: MJG, DAL and LACM all contributed to \nthe planning of this project. MJG conducted the analysis and wrote the article with \nsupervision and support from DAL and LACM. The corresponding author attests that \nall listed authors meet authorship criteria and that no others meeting the criteria have \nbeen omitted. \nPatient and public involvement reporting: Neither patients or the public were \ninvolved at any stage of this study. \nCopyright/license for publication: The Corresponding Author has the right to grant \non behalf of all authors and does grant on behalf of all authors, a worldwide \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nlicence to the Publishers and its licensees in perpetuity, in all forms, formats and \nmedia (whether known now or created in the future), to i) publish, reproduce, \ndistribute, display and store the Contribution, ii) translate the Contribution into other \nlanguages, create adaptations, reprints, include within collections and create \nsummaries, extracts and/or, abstracts of the Contribution, iii) create any other \nderivative work(s) based on the Contribution, iv) to exploit all subsidiary rights in the \nContribution, v) the inclusion of electronic links from the Contribution to third party \nmaterial where-ever it may be located; and, vi) licence any third party to do any or all \nof the above. \nThis work carries a Creative Commons Attribution (CC BY 4.0) license.  \nCompeting interests declaration: All authors have completed the ICMJE uniform \ndisclosure form at http://www.icmje.org/disclosure-of-interest/. DAL has received \nsupport from Roche Diagnostics and Medtronic Ltd for biomarker research unrelated \nto this paper. MJG and LACM declare no support from any organisation for the \nsubmitted work; no financial relationships with any organisations that might have an \ninterest in the submitted work in the previous three years; no other relationships or \nactivities that could appear to have influenced the submitted work. \nData sharing statement: All data is available on request from the UK Biobank or \n23andMe. The full GWAS summary statistics for the 23andMe discovery data set will \nbe made available through 23andMe to qualified researchers under an agreement \nwith 23andMe that protects the privacy of the 23andMe participants. Please visit \nhttps://research.23andme.com/collaborate/#dataset-access/ for more information \nand to apply to access the data. \nEthical approval: The data collection in UK Biobank was approved by the NHS \nNational Research Ethics Service (ref 11/NW/0382). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nTransparency statement: The lead author affirms that the manuscript is an honest, \naccurate, and transparent account of the study being reported; that no important \naspects of the study have been omitted; and that any discrepancies from the study \nas originally planned have been explained. \nSupplemental material: This content has been supplied by the author(s). It has not \nbeen vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-\nreviewed. Any opinions or recommendations discussed are solely those of the \nauthor(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility \narising from any reliance placed on the content. Where the content includes any \ntranslated material, BMJ does not warrant the accuracy and reliability of the \ntranslations (including but not limited to local regulations, clinical guidelines, \nterminology, drug names and drug dosages), and is not responsible for any error \nand/or omissions arising from translation and adaptation or otherwise. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \n \nFigure 1 \nFlow chart of participant inclusion. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 2  \nProportion of potential causal effects of insomnia on outcomes within different \ncategories. \n \nn is the total number of outcomes in the category. Supplementary Table S3 gives the \ncategory for each outcome. Results shown in this figure are also provided in \nSupplementary Table S4. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 3 \nProportion of potential causal effects of insomnia on outcomes within different \nmental health subcategories.  \n \nn is the total number of outcomes in the category. Supplementary Table S3 gives the \nsubcategory for each outcome. Results shown in this figure are also provided in \nSupplementary Table S5. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 4 \nProportion of potential causal effects of insomnia on outcomes within different \nphysical health subcategories.  \n \nn is the total number of outcomes in the category. Supplementary Table S3 gives the \nsubcategory for each outcome. Results shown in this figure are also provided in \nSupplementary Table S5. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 5 \nFlow chart of GWAS inclusion for follow-up. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 6  \nTwo sample MR results of the effect (odds ratio), comparing self-reported \ninsomnia cases versus non-cases (see Supplementary Text for 23andMe \ninsomnia definition), for binary outcomes. \n \n*GWAS has overlap with UK Biobank or 23andMe (see Supplementary Table S7 for \nexact percentage). \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint \n\n \n \nFigure 7 \nTwo sample MR results of the effect (mean difference), comparing self-\nreported insomnia cases versus non-cases (see Supplementary Text for \n23andMe insomnia definition), for continuous outcomes. \n \n*GWAS has overlap with UK Biobank or 23andMe (see Supplementary Table S7 for \nexact percentage). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted January 14, 2022. ; https://doi.org/10.1101/2022.01.11.22269005doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}