Received: 5 January 2024 Accepted: 18 January 2024

DOI: 10.1111/gbb.12885

REVIEW ARTICLE

Neuropsychiatric disorders, chronotype and sleep: A narrative

review of GWAS findings and the application of Mendelian
randomization to investigate causal relationships

Shane Crinion 1,2 | Derek W. Morris 1 | Lorna M. Lopez 2

1
Centre for Neuroimaging, Cognition and
Genomics, School of Biological and Chemical Abstract
Sciences, University of Galway, Galway,
Genome-wide association studies (GWAS) have been important for characterizing
Ireland
2
Department of Biology, Maynooth University,
the genetic component and enhancing our understanding of the biological aetiology
Maynooth, Ireland of both neuropsychiatric disorders and sleep-related phenotypes such as chronotype,

Correspondence which is our preference for morning or evening time. Mendelian randomization
Derek W. Morris, School of Biological and (MR) is a post-GWAS analysis that is used to infer causal relationships between
Chemical Sciences, University of Galway,
Galway, Ireland.
potential risk factors and outcomes. MR uses genetic variants as instrumental vari-
Email: [email protected] ants for exposures to study the effect on outcomes. This review details the main

Funding information
results from GWAS of neuropsychiatric disorders and sleep-related phenotypes, and
European Research Council, Grant/Award the application of MR to investigate their bidirectional relationship. The main results
Number: 950010; Science Foundation Ireland,
Grant/Award Numbers: 15/SIRG/3324, 18/
from MR studies of neuropsychiatric disorders and sleep-related phenotypes are
CRT/6214 summarized. These MR studies have identified 37 causal relationships between neu-
ropsychiatric disorders and sleep-related phenotypes. MR studies identified evidence
of a causal role for five neuropsychiatric disorders and symptoms (attention deficit
hyperactivity disorder, bipolar disorder, depressive symptoms, major depressive dis-
order and schizophrenia) on sleep-related phenotypes and evidence of a causal role
for five sleep-related phenotypes (daytime napping, insomnia, morning person, long
sleep duration and sleep duration) on risk for neuropsychiatric disorders. These MR
results show a bidirectional relationship between neuropsychiatric disorders and
sleep-related phenotypes and identify potential risk factors for follow-up studies.

KEYWORDS
chronotype, genome-wide association studies, Mendelian randomization, neuropsychiatric
disorders, sleep

1 | I N T RO DU CT I O N childhood-onset conditions of autism spectrum disorder (ASD) and

attention deficit hyperactivity disorder (ADHD). These disorders were
Neuropsychiatric disorders are a complex group of disorders affecting the original disorders identified for gene discovery research by the
brain function, behaviour and cognition.1 The major disorders include Psychiatric Genomics Consortium (PGC).2
the adolescent and adult-onset disorders of bipolar disorder (BD), There is wide variability in the characteristics of neuropsychiatric
major depressive disorder (MDD) and schizophrenia (SZ), and the disorders which makes the identification of environmental and genetic

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2024 The Authors. Genes, Brain and Behavior published by International Behavioural and Neural Genetics Society and John Wiley & Sons Ltd.

Genes, Brain and Behavior. 2024;23:e12885. wileyonlinelibrary.com/journal/gbb 1 of 13

https://doi.org/10.1111/gbb.12885
2 of 13 CRINION ET AL.

risk factors difficult. The diversity in cognitive profile and behavioural Establishing a causal relationship between chronotype and neuropsy-
characteristics between BD, MDD and SZ3 suggests the presence of a chiatric disorders via MR could identify sleep-related phenotypes as
neurodevelopmental continuum with variations in cognitive impair- modifiable risk factors for treating the disabling features of neuropsy-
ment, negative symptoms, positive symptoms and mood disturbance.4 chiatric disorders. Further understanding of the complex relationship
ASD and ADHD have an earlier clinical onset and a wide range of clin- between neuropsychiatric disorders and environmental exposures will
ical features, including global or local impairment of social skills, learn- define new approaches for interventions.
3,4
ing and executive functions. In this narrative review, we highlight the insights gained from
Circadian rhythms are patterns in physiology and behaviour that genetic investigations using genome-wide association studies (GWAS)
recur with 24 h frequency and are driven by endogenous mechanisms of neuropsychiatric disorders, chronotype and other sleep-related
and entrained to environmental time cues.5 Overall regulation of the phenotypes. We review the application of Mendelian randomization
human circadian rhythms is maintained by a central clock located in (MR) to explore the potential causal relationships between neuropsy-
the suprachiasmatic nuclei (SCN), a pair of small nuclei containing chiatric disorders and sleep-related phenotypes. MR uses single nucle-
20,000 neurons located in the anterior hypothalamus. The SCN coor- otide polymorphisms (SNPs) from GWAS as instrumental variants for
dinates or ‘orchestrates’ circadian rhythms at a systemic level through exposures to study the effect on outcomes.
neural and endocrine mechanisms. The molecular clock is the core ele- The approach used for this narrative review was to identify and
ment of circadian rhythms that is present in virtually every cell, and summarize evidence for the genetic component of neuropsychiatric dis-
that can even be reproduced in vitro in cell culture. The molecular orders and chronotype while also highlighting existing evidence for the
clock confers 24 h rhythmicity on physiology and behaviour, but interrelationship between neuropsychiatric disorders and sleep traits.
endogenous regulation of most of these rhythms will fail in the While we aimed to inform on the published studies that are avail-
absence of SCN input. Circadian rhythm alterations can be assessed able that provide a broad perspective of the genetic components of
by SCN-regulated responses such as body heat temperature,6 sleep– these phenotypes, we did not apply a systematic review approach. A
wake activity and self-reported measures such as chronotype.7–9 systematic review has a specific clinical question and typically involves
Chronotype is thought to be a behavioural indicator of an individ- meta-analysis, which was not possible here because we explored mul-
ual's circadian rhythm for example, being a morning person (morning- tiple phenotypes within the two categories of neuropsychiatric disor-
ness) means having a preference for sleep and activities earlier in the ders and sleep-related phenotypes. Although this review does not
24 h day, while evening chronotypes prefer later sleep, wake and include the methodological requirements of a systematic review, this
7
activity time. Chronotype is a sleep-related phenotype that has been narrative review remained systematic and applied the Scale for the
used to characterize the function of the human circadian clock. Assessment of Narrative Review Articles (SANRA) methodology,
Chronotype is influenced by age, biological sex and environmental which outlines six items for quality assessment of narrative reviews.24
factors, but it also has a strong genetic component and is one of the
sleep-related phenotypes with the largest genetic studies available to
investigate its biology.7 Eveningness, describing someone's propensity 1.1 | Genetics of neuropsychiatric disorders
for waking and sleeping later, has been linked to increased risk for
neuropsychiatric disorders10 and circadian rhythm disruption,11 as Six neuropsychiatric disorders (ADHD, ASD, BD, Insomnia, MDD, SZ)
estimated by disruption of diurnal patterns of rest-activity such were selected for review. GWAS have identified SNPs and led to
as sleep–wake cycle and core body temperature variation.12 improved understanding of the genetic risk factors for neuropsychiat-
Disruption of circadian timing is a feature of most neuropsychiat- ric disorders and, as sample sizes continue to grow in GWAS meta-
ric disorders, manifesting as disturbed sleep and diurnal changes in analyses by the PGC and other groups, will likely have improved
13
behaviour. Sleep and circadian rhythm disruption is present at all power to detect more risk variants. Therefore, we included in this
stages of BD.14 Severe BD type I cases have showed lower circadian review the major neuropsychiatric disorders where a large-scale,
rhythm activity than other BD cases.15 MDD is up to 40% more likely sufficiently-powered GWAS had been performed.
in night shift workers versus day workers.16 Sleep abnormalities occur While not one of the major neuropsychiatric disorders, we also
17
in up to 80% of SZ patients. Between 25%–55% of children with included insomnia, which affects 10% of non-psychiatric individuals25
18
ADHD have sleep disturbances and 50%–80% of children with ASD and is associated with several neuropsychiatric disorders.3
19
have sleep problems. For comparison, sleep disturbances affect GWAS for neuropsychiatric disorders and accompanying down-
25%–30% of adults worldwide.10 Disruption to overlapping biological stream analyses have characterized the genetic component and
pathways and mechanisms has also been discovered in studies on the showed their implicated cell types and biological processes, which are
comorbidity of circadian rhythm disruption and neuropsychiatric summarized in Table 1. The number of SNPs reported as
disorders.20 genome-wide significant loci for the selected neuropsychiatric disor-
Life-long disability and the severity of neuropsychiatric disorders ders ranged from 5 (ASD) to 554 (Insomnia). Calculations of SNP
21
can be arguably reduced through early intervention. However, there based heritability (h2SNP), which represents the proportion of pheno-
are currently few approaches available due to the complex and poorly typic variance due to all measured SNPs and can be estimated from
understood neuropathology of neuropsychiatric disorders.22,23 GWAS data, ranges from 7% (insomnia) and up to 24% (SZ).
CRINION ET AL. 3 of 13

TABLE 1 Neuropsychiatric disorder GWAS: Main findings.

GWAS Enriched tissue and cell Enriched biological

Phenotype Cases Controls locia h2SNP types processes Biological insights References
ADHD 20,183 35,191 12 0.22 CNS regulatory elements. Dopamine receptor LDSR: " MDD, depressive Demontis
binding. symptoms; # smoking et al26
Excitatory Synapse. and subjective well-
being.
ASD 18,381 27,969 5 0.12 Human neocortex modules Corticogenesis LDSR: " SZ, EA, MDD & Grove
M16 and M17. regulatory ADHD; # chronotype & et al27
H3K4me1 histone marks. elements. subjective well-being.
CNS. Foetal PRS: Strong heterogeneity
Developing brain, germinal corticogenesis. found in ASD subgroups
matrix, cortex-derived split by diagnostic class
neurospheres, and ESC- and cognitive
derived neurons. phenotypes.
BD 41,917 371,549 30 0.19 Hippocampal pyramidal Neuronal processes. LDSR: " correlation with Mullins
neurons. Synaptic functioning SZ, MDD, Anorexia, et al28
Interneurons of the Calcium signalling ADHD & ASD.
prefrontal cortex and Neurogenesis Drug targets: Enrichment
hippocampus. of psycholeptics, calcium
Excitatory and inhibitory channel blockers,
neurons. antiepileptics and
general anaesthetics
(HTR6, MCHR1, DCLK3
and FURIN).
Insomnia 593,724 1,771,286 554 0.07 Tissues of cerebellar Synaptic LDSR: " cardiovascular, Watanabe
hemisphere, cerebellum, organization, metabolic and et al29
frontal cortex BA9 and transmission and psychiatric traits.
anterior cingulate cortex signalling. Colocalization: results
BA24. Behaviour. indicated 2 distinct loci-
Lateral geniculate nucleus, trait clusters with
Habenula, ventral metabolic and
pallidum and anterior psychiatric traits.
pretectal nucleus.
MDD 246,363 561,190 102 0.09 CNS. Behaviour, cognition LDSR: " SZ, BD; # college Howard
Anterior cingulate cortex, and synaptic completion et al30
frontal cortex and cortex transmission. Drug targets: DRD2 and
brain regions and Mood modulation. NRG1, targets of
neuron brain cells. Emotion processing. antipsychotic and
Skeletal muscle tissues. antidepressant drugs.
SZ 69,369 236,642 294 0.24 Cortical inhibitory Neuronal excitability, Fine-mapping: highlights Trubetskoy
interneurons. development, and convergence of rare and et al31
Excitatory neurons from structure. common variants in 4
cerebral cortex and Synaptic organisation specific genes (GRIN2A,
hippocampus (pyramidal and differentiation. SP4, STAG1 and
and granule cells). Modulation of FAM120A).
Glutamatergic neurons in chemical Drug targets: ACE, a target
the cortex, amygdala, transmission. for antihypertensive
and hippocampus. Postsynaptic drugs.
processes
implicated in risk.

Abbreviations: ADHD, attention deficit hyperactivity disorder; ASD, autism spectrum disorder; BD, bipolar disorder; CNS, central nervous system; EA,
educational attainment; ESC, embryonic stem cell; GWAS, genome-wide association studies; LDSR, linkage disequilibrium score regression; MDD, Major
Depressive Disorder; MR, Mendelian randomisation; PRS, polygenic risk score; SZ, Schizophrenia.
a
Independent genome-wide significant loci ( p < 5  108).
4 of 13 CRINION ET AL.

The findings from gene set enrichment analyses (GSEA) indicate linked to PDE11A variants, a potential target gene for mood stabiliza-
that the frontal cortex, prefrontal cortex and hippocampus play an tion in neuropsychiatric disorders.37 PDE11A SNPs associated with
important role in these neuropsychiatric disorders. GSEA has also sleep duration were previously linked to schizophrenia and migraines.8
showed that genes associated with a neuropsychiatric disorder are Gene set analysis has also been used to identify an enrichment of
significantly overrepresented in biological processes involved in devel- genes expressed in the cerebellum for all sleep measurements
opment, function and signalling of neurons and synapses. Genetic cor- (Table 2), a brain region linked to neuropsychiatric disorders that,
relation analyses have showed significant positive associations when dysfunctional, can lead to changes in sleep–wake cycle.
between numerous neuropsychiatric disorders and cognitive, beha- Genetic correlation has also been used to show significant genetic
vioural and psychiatric traits (Table 1). overlap between sleep-related phenotypes and neuropsychiatric dis-
orders. Jones et al7 reports that morningness is positively correlated
with subjective well-being and negatively correlated with SZ, depres-
1.2 | Genetics of sleep-related phenotypes sive symptoms and MDD, providing further evidence of shared biol-
ogy. Figure 1 shows up-to-date results obtained using the GWASatlas
Table 2 summarizes the main findings from GWAS and post-GWAS repository38 of GWAS results. These data firstly indicate strong
analyses of sleep-related phenotypes. Chronotype was selected as the genetic correlations within the group of neuropsychiatric disorders as
main phenotype to study sleep-related phenotypes due to the avail- well as strong genetic correlations within sleep-related phenotypes.
ability of the largest samples for genetic studies. A GWAS of 697,828 Across these two groups of phenotypes, morningness was found to
individuals was performed and detected 351 genome-wide significant be negatively correlated with ASD and SZ but M10 timing, the mid-
7
(GWS) loci associated with self-reported chronotype. These SNPs point of most active 10 hours, was found to be positively correlated
were found in genes that control circadian regulation, cAMP, gluta- with ASD and SZ. Sleep duration was found to be positively
mate and insulin signalling pathways. correlated with BD and SZ and negatively correlated with insomnia.
Most GWAS for sleep-related phenotypes have used self- However, while these correlations are useful for identifying traits that
reported measures; however, recent GWAS have been performed on influence one another, they do not show the causal associations
accelerometer-recorded sleep measurements. Jones et al published a between traits.
GWAS of 85,670 individuals using accelerometer measurements that
detected 47 GWS loci associated with sleep quality, sleep quantity
and sleep timing (Table 2).8 Sleep quality was estimated using mea- 1.4 | Mendelian randomization for causal inference
surements for (1) sleep efficiency (sleep duration divided by the time using genetic data
between the start and end of the first and last nocturnal inactivity
period) and (2) the number of nocturnal sleep episodes. Sleep timing 1.4.1 | What is MR?
was estimated using (1) midpoint of least active 5 h (L5), (2) midpoint
of most active 10 h (M10) and (3) sleep midpoint (Table 2). Number of MR is a post-GWAS analysis that can provide evidence to support or
nocturnal sleep episodes was found to have the largest genetic com- reject hypotheses of causal relationships between an exposure trait
ponent with SNPs accounting for 22.3% of phenotypic variance and risk for a disease outcome.39
(Table 2). An additional GWAS of 446,118 individuals performed by
Dashti et al32 identified 78 GWS loci associated with self-reported
sleep duration and found SNPs associated with pathways such as 1.4.2 | What are IVs?
dopamine binding and neurotransmission and plasticity (Table 2).
Trait-associated genetic variants, which are strongly associated with
an exposure trait, are employed as instrumental variants (IVs) to repre-
1.3 | Genetic relationship between sent the exposure and assess it as a potentially causal risk factor for
neuropsychiatric disorders and sleep-related the outcome. For polygenic traits, the IVs selected are typically inde-
phenotypes pendent SNPs identified through GWAS.

There is genetic evidence of a bidirectional relationship between neu-

ropsychiatric disorders and circadian rhythm.33 Disruption to overlap- 1.4.3 | How does MR compare with randomized
ping biological pathways and mechanisms has been discovered in controlled trials?
studies on the comorbidity of circadian rhythm disruption and neuro-
psychiatric disorders.20 Other evidence includes a study using animal The MR concept and design is outlined and compared with random-
models that induced mania-like behaviours through mutations in the ized controlled trials (RCTs) in Figure 2A. Conditional on some
34
Clock gene and overlapping mutations in the CLOCK gene have been assumptions, MR is similar to RCTs, which are performed to study the
linked to both neuropsychiatric disorders and circadian rhythm disrup- effect of various therapies, exposures, or behaviours on disease risk.
tion.35,36 Increased sleep duration and efficiency have also been During an RCT, participants are randomly assigned to one of two
CRINION ET AL. 5 of 13

TABLE 2 Chronotype and sleep GWAS: Main findings.

GWAS H2 Enriched tissue and Enriched biological

Phenotype Type Individuals locia SNP cell types processes Biological insightsb References
Chronotype Self-report 697,828 351 0.14 SCN genes Circadian rhythm and Fine-mapping: Jones
Circadian gene and circadian clock Identified 10 likely et al8A
circadian-implicated pathways CNS & causal variants, link
genes synapses, brain development to insulin secretion.
axons and dendrites Neurogenesis. NMDA pathway
retinal tissue Behavioural pathways genes NRXN1 and
Responses to internal RELN influence the
& external stimuli risk of SZ.
Metabolism of cyclic
nucleotides
G-protein signalling
and activation.
NMDA glutamate
signalling pathway
Sleep Measured 85,502 5 0.13 Cerebellum None reported. Fine-mapping: likely Jones
efficiency Hippocampus causal variant in et al8B
hippocampus
expressed PDE11A.
Overlap in MEIS1,
linked to restless
leg syndrome/
insomnia.
No. sleep Measured 85,502 21 0.22 Cerebellum Serotonin metabolic APOE ε4 allele variant, Jones
episode process associated with et al8B
late-onset
Alzheimer's and
cognitive decline—
increase in allele
frequency by age.
CLUAP1 variant,
associated with
photoreceptor
maintenance.
L5 Measured 85,723 6 0.12 Cerebellum None reported. APOE ε4 risk allele Jones
identified. et al8B
LDSR: High genetic
correlation with
chronotype.
M10 Measured 85,830 1 0.09 Cerebellum None reported LDSR: High genetic Jones
correlation with et al8B
chronotype.
Sleep duration Self-reported 446,118 78 0.09 Cerebellum, Cortex, Dopamine binding, LDSR: Negative Dashti
Hippocampus, mechanosensory genetic correlation et al32
Hypothalamus behaviour, striatum with insomnia and
development positive genetic
correlation with SZ
and BD.
Sleep midpoint Measured 85,502 1 0.10 Cerebellum None reported APOE ε4 risk allele Jones
identified. et al8B
LDSR: High genetic
correlation with
chronotype.

Abbreviations: GWAS, genome-wide association studies; LDSR, linkage disequilibrium score regression; L5, Midpoint of least active 5 h; M10, Midpoint of
most active 10 h; SCN, suprachiasmatic nucleus.
a
Independent genome-wide significant loci ( p < 5  108).
b
" represents an LDSR positive genetic correlation between phenotypes and # represents an LDSR negative genetic correlation between phenotypes.
6 of 13 CRINION ET AL.

F I G U R E 1 Genetic correlation results

indicating shared biology of
neuropsychiatric disorders and sleep-
related phenotypes. Symmetric heatplot
displaying genetic correlations where
those denoted by ‘*’ are significant at
p < 0.05 after Bonferroni correction for all
tests. L5 timing is the midpoint of the
least active 5 h of each day. M10 timing is
the timing of the most active 10 h of each
day. This was generated using genome-
wide association studies (GWAS) data
from European ancestry individuals from
GWASatlas. GWASatlas ID and year of
publication (brackets) are included for
each phenotype. Results are clustered by
degree of genetic correlation.

study groups, such as an exposure (treatment) group and a control be valid: (1) IVs are robustly associated with the risk factor, hence usu-
group. The results are compared between the two groups and any sta- ally only SNPs that are independent and GWS are used as IVs; the rel-
tistically significant difference is deemed to be due to the assigned evance assumption. (2) IVs share no common causes with the
exposure or treatment. In MR analysis, individuals with trait associ- outcome; the independence assumption (sometimes known as
ated alleles are roughly analogous to the study group assigned to the marginal exchangeability assumption). (3) IVs have no association
39
treatment in a randomized trial (Figure 2A). In a true randomized with the outcome that is not mediated through the risk factor; the
trial, individuals are assigned at random to treatment and control exclusion restriction assumption.
groups. However, MR exploits the random assignment of disease Potential sources of bias include weak instruments, horizontal
associated alleles (conditional on parental genotype) at conception to pleiotropy and reverse causality.
individuals to assign individuals to study groups. Violation of the relevance assumption (Assumption 1) can intro-
duce weak instrument bias, a phenomenon causing large statistical
variability and bias in the MR estimate when variants only weakly
1.5 | IV selection process associated with the trait are used as instruments.42 The values used
to measure instrument strength include R2, a measure that estimates
IVs are selected based on their known biological function or statistical the variance of the trait explained by the variant(s), and the F-statistic,
association with the exposure trait from GWAS and must meet the IV which accounts for R2, sample size and the number of IVs. In
assumptions.41 If there is a causal relationship between the exposure Figure 2B, weak instrument bias is introduced if the link between
and the outcome, the IVs that are associated with the exposure will genetic variants and the morning chronotype is not strong, for exam-
be associated with the outcome in a proportional way—the effect ple, if the variants used as IVs are not GWS for morning chronotype.
sizes can be much smaller, but a SNP with a large effect on exposure Violation of the independence assumption (Assumption 2) can
will have a large effect on outcome, and similarly for small effect. introduce bias through confounders.43 Confounding refers to the case
where a characteristic influences the exposure and outcome through
a ‘backdoor pathway’ between the exposure and outcome.44 This can
1.6 | Assumptions occur when the genetic variants used as IVs influence both the con-
founders and the outcome, which could exaggerate or attenuate the
Figure 2B outlines a pertinent example for an MR investigation of the causal effect estimate. This scenario is analogous to horizontal pleiot-
causal role of the morning chronotype on SZ risk. There are three core ropy, whereby an IV influences multiple traits and influences the out-
IV assumptions that generally must hold in order for the MR study to come through an alternative biological pathway that is beyond the
CRINION ET AL. 7 of 13

F I G U R E 2 Mendelian randomisation concept and design. (A) Comparison of randomized controlled trial and MR study design. During a
randomized controlled trial (RCT), study participants are randomly assorted during study design into two groups and an exposure (e.g., a
treatment) is applied to one group to understand the extent to which it influences risk for the outcome. Outcomes (e.g., depression) are compared
between both groups and any difference is considered to be due to the exposure treatment. Some treatments such as a set sleep duration are
impractical and unethical. Alternatively, MR can be used for any genetically-based traits. During MR, SNP alleles are assigned as instrumental
variants (IVs) to assign groups. One group represents the exposure-risk allele (e.g. carriers of alleles associated with short sleep duration) and one
group with the other non-risk allele. Outcomes are compared between both groups and any difference is considered to be due to the genetically-
based exposure. (B) Schematic diagram of MR study design and assumptions. During an MR analysis, the effect estimates for genetic instruments
are extracted from genome-wide association studies data and used to approximate the effect of the exposure on the outcome. The study design
of MR is dependent on three core assumptions—(1) the genetic instruments are robustly associated with the exposure, (2) genetic instruments
have no association with outcome that is not mediated through the risk factor and (3) genetic instruments are not directly associated with the
outcome.40

examined relationship. Figure 2B illustrates how bias from horizontal depicts a scenario whereby this assumption is void and IVs for morn-
pleiotropy can occur if genetic variants linked to the morning chrono- ing chronotype are also linked to SZ. Bias can be investigated by using
type also influence confounders (e.g., work hours, drugs, diet). MR methods developed to detect horizontal pleiotropy such as MR-
Bias from horizontal pleiotropy can also occur through violation Egger,45 weighted median46 and mode-based MR.47 Table S1 provides
of the exclusion restriction assumption (Assumption 3) whereby the definitions from the MR dictionary of each of these MR methods.
genetic variants selected as IVs are associated with the outcome but Beyond the core IV assumptions, the researcher should also examine
not through the hypothesized exposure-outcome pathway. Figure 2B whether reverse causality is observable. Should valid IVs be available
8 of 13 CRINION ET AL.

for both the exposure and the outcome, a bidirectional MR should be causal effects and the MR sensitivity tests. The search strategy
applied to determine whether the outcome is influencing the yielded 30 original search results, of which 15 included relevant MR
exposure. analyses that investigated the causal relationship between neuropsy-
There are several methodological approaches to MR but two- chiatric disorders and sleep traits. Table S2 contains all significant and
sample MR (2SMR) is often the most convenient to apply48 and was non-significant MR analyses and lists the exposure and outcome
the method used for all the studies reviewed here. 2SMR uses GWAS GWAS data for each.
summary statistics that are often publicly available and can increase The methods to address IV assumptions and bias, such as sample
power and improve predictive ability. IVs are selected using the overlap and reverse causality, for each of these studies are outlined in
GWAS data for the exposure and then the test statistics for the corre- Table S3. All studies used IVs identified from GWAS data. Generally,
sponding IVs are extracted from the outcome GWAS data. There are the p-value threshold of <5  108 was used to assess strength of
three additional assumptions for 2SMR analysis: (1) the two GWAS association; however, the p-value was reduced in several studies to
samples are non-overlapping, (2) the two samples are from the same assess reverse causality. The number of IVs used in the MR analyses
underlying population and (3) genetic instruments are harmonized included in this review ranged from 6 to 340. There were several MR
between both samples so that the effect or risk allele and the other methods used in these studies to infer causality including inverse-
44
non-risk allele are concordant between the two GWAS datasets. variance weighted (IVW), penalized-weighted median (PMW), MR-
Egger and MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO)
methods. The results are outlined in Figure 3 and classified according
1.7 | Application of MR to neuropsychiatric to the type of exposure (neuropsychiatric disorders or sleep-related
disorders and sleep-related phenotypes phenotypes).
From these 15 studies, 37 gave significant MR results between
Epidemiological studies of neuropsychiatric disorders have been used neuropsychiatric disorders and sleep-related phenotypes which are
to evaluate the contribution of exposures such as genetics, environ- interpreted as consistent with a causal relationship. Of these 37 rela-
mental factors and prenatal factors. While epidemiological studies tionships, a total of 15 were identified where a neuropsychiatric disor-
have been useful for understanding the aetiology of neuropsychiatric der was found to be associated with a sleep-related phenotype and a
disorders, it is not ethically possible to perform interventional total of 22 were found where a sleep-related phenotype was found to
methods such as RCTs, due to the exposure of participants to poten- be associated with a neuropsychiatric disorder. Figure 3 features
tially harmful risk factors that could initiate or exacerbate negative results from the MR method that the authors used to conclude as
features of neuropsychiatric disorders. However, due to the increasing showing evidence for a significant causal relationship. However, each
number of GWAS studies and genetic correlation studies, it has been MR method has advantages and limitations and interpretations of a
possible to identify potential risk factors using MR research to assess significant result can vary.
causality.
Numerous MR studies have been published that report evidence
for associations between neuropsychiatric disorders and sleep-related 1.7.1 | Neuropsychiatric disorders as an exposure
phenotypes. These studies used genetically proxied exposures to for sleep-related phenotypes
assess causality for outcomes of interest. Several of these studies
used overlapping GWAS datasets (e.g., UK Biobank, Psychiatric Geno- The results from MR studies of neuropsychiatric disorders as expo-
mics Consortium, 23andMe). However, there is no included study that sures found that five neuropsychiatric disorders and symptoms
uses identical MR methodologies, GWAS datasets and IV selection (ADHD, BD, depressive symptoms, MDD and SZ) were found to be
process. All these MR studies used samples of European ancestry. We associated with sleep-related phenotypes. ADHD was found to
searched PubMed up to 25th April 2023 for MR studies investigating be associated with lower probability of being a morning person and
the association of sleep phenotypes with neuropsychiatric disorders causal for long sleep duration.49 However, ADHD was conversely
using Mendelian randomization. We used the search term [(Mendelian reported as associated with reduced sleep duration by Sun et al.50
randomization OR Mendelian randomization) AND (neuropsychiatric BD was found to be associated with increased sleep duration and
OR psychiatric) AND (autism OR ASD OR schizophrenia OR depres- lowered probability of being a morning person.28 Depressive symp-
sive OR depression OR ADHD OR attention deficit OR bipolar OR toms were found to be associated with increased insomnia risk.25
BD) AND (chronotype OR sleep)]. We also screened the relevant MDD was found to increase risk for daytime napping,51 insomnia,52
GWAS for each of the included neuropsychiatric disorders for Mende- sleep apnoea53 and sleep disorders54 and associated with short sleep
lian randomization analysis performed that investigated the link with duration.49 MDD was reported as associated with insomnia in three
sleep or chronotype. We extracted the exposure, outcome, genetic separate studies25,52,55,56; however there was wide variation in the
instrument, MR design (one sample or two-sample and the relevant reported OR (1.06–2.23) but concordant direction of effect. Finally,
population) and the type of MR that was performed to investigate the SZ was found to be associated with longer sleep duration.25
CRINION ET AL. 9 of 13

F I G U R E 3 Mendelian randomisation analysis for neuropsychiatric disorders and sleep-related phenotypes. Results outlined in this figure
were compiled from all MR analyses which investigate causal relationships between neuropsychiatric disorders and sleep-related phenotypes.
Bold headers indicate the exposure category. Column ‘SNPs’ indicates the number of genetic instruments used for the exposure. Each point
represents the odds ratio and confidence intervals. Where necessary, beta values were converted to odds ratios. The size of each point is
proportional to the inverse standard error, meaning that larger points are indicative of more precise estimates. The p-value is as reported in the
original study. Odds ratios below 1 indicate that the exposure/risk factor is protective for the outcome, while odds ratios above 1 indicate that
the exposure/risk factor increases risk for the outcome.

1.7.2 | Sleep-related phenotypes as an exposure for be associated with increased risk for SZ32 and sleep duration was
neuropsychiatric disorders found to increase risk for BD.28,50

The results from MR studies of sleep-related phenotypes as expo-

sures found that four sleep-related phenotypes (daytime napping, 2 | DI SCU SSION
insomnia, long sleep duration and sleep duration) were associated
with increased risk of neuropsychiatric disorders and morningness Here, we have reviewed GWAS for neuropsychiatric disorders and
was found to be associated with lower risk of neuropsychiatric disor- sleep-related phenotypes with a particular focus on MR to examine
ders (Figure 3). Daytime napping was found to be associated with the potential causal relationships between neuropsychiatric disorders
increased risk for ADHD49 and MDD.51 Carpena et al also report that and sleep-related phenotypes. GWAS has led to the identification of
short sleep duration is associated with ADHD, however this result hundreds of SNPs that are associated with neuropsychiatric disorders
was not displayed here due to relatively weak evidence of causality and sleep-related phenotypes. Post-GWAS analyses using various
with very large estimated effect size across MR methods, and inverse methods has led to characterization of putative risk genes and biologi-
direction of effect across MR methods. Insomnia was found to be cal pathways.
associated with increased risk for ADHD,25,49,50 ASD,57 BD,57,58 The MR studies outlined in this review have identified 37 poten-
25,59 25,52,55,56
depressive symptoms and MDD. Being a morning per- tially causal relationships between neuropsychiatric disorders and
son was found to be associated with decreased risk of depressive sleep-related phenotypes. All the included MR studies performed
symptoms,60 MDD50,61 and SZ.50 Long sleep duration was found to numerous MR methodologies to assess causal effect and sensitivity
10 of 13 CRINION ET AL.

tests to assess bias. Five neuropsychiatric disorders and symptoms evidence of a causal effect is reported despite lack of consistency in
(ADHD, BD, depressive symptoms, MDD and SZ) were found to influ- effect estimates across MR and sensitivity tests and, in some cases,
ence risk for sleep-related phenotypes while five sleep-related traits inverse trends in the main MR and sensitivity tests.
(daytime napping, insomnia, morning person, long sleep duration and Variations in MR analysis exist that may enhance studies by
sleep duration) were found to be associated with neuropsychiatric dis- reducing bias and integrating multi-omic data. For example, an
orders. Insomnia was found to be associated with ADHD, ASD, BD, approach has been developed to differentiate genetic instruments
depressive symptoms and MDD. Insomnia was identified as having a based on their predicted mechanism.63 The approach can then iden-
bidirectional causal relationship with MDD, highlighting the intricate tify when a risk factor is causing multiple biochemical changes that
relationship between neuropsychiatric disorders and sleep-related influence risk for an outcome. Implementation of this method could
phenotypes, by which one can exacerbate the other. One other be useful for neuropsychiatric disorders, where many of the GWS loci
potential bidirectional causal relationship was identified (BD and sleep used as genetic instruments have both protective and causal effects.
duration). It was not possible to test for a bidirectional causal relation- As many GWS loci can also be eQTLs, the incorporation of expression
ship between all combinations of phenotypes as some lacked suffi- and gene regulation data in an MR framework can be used to identify
cient GWS SNPs as IVs. Therefore, the true nature of the causal causal gene-trait associations and complement GWAS results and
relationship between some phenotypes remains to be determined. identify priority genes. Transcriptome-wide summary statistics-based
Those analyses will require further growth of GWAS and an increased MR approach (TWMR) uses gene expression changes as an exposure
number of SNPs that can be used as IVs. to understand its effect on an outcome and has led to 3913 novel
Several potential causal relationships reported in this review were gene-trait associations, many of which were later found in larger
reported in more than one MR study. These studies reported variable GWAS.64 TWMR can incorporate multi-omic data, such as methyla-
effect size estimates but had consistent concordance in the predicted tion QTLs, to find functional changes that are causal for an outcome.
direction of effect, with the caveat that not all studies may have used In conclusion, GWAS and MR have been useful for understanding
independent samples. Therefore, it is important to recognize that MR the complex, bidirectional relationship between neuropsychiatric dis-
research can be accurate in identifying risk factors and predicting the orders and sleep. The phenotypic variance of neuropsychiatric disor-
direction of effect but the effect size is representative of lifetime con- ders has made it difficult to generate knowledge on the core biology
tribution to risk and does not represent the effect size that an inter- and has hindered the development of effective treatments for these
vention would cause at a specific time.44 Exposures that are disorders. MR enables a powerful strategy that focuses on under-
modifiable risk factors have clear clinical relevance and can influence standing the relationship between circadian rhythm and neuropsychi-
treatment, if follow-up studies validate these associations. Other atric disorders without requiring a complete understanding of the
exposures (e.g., SZ) give insight into the biological aetiology of the underlying biological origins of neuropsychiatric disorders. In this
associated outcomes. sense, MR is an epidemiological tool that employs genetic data. MR
These results provide evidence that sleep and circadian rhythm studies can also help us in navigating drug development targets and
can contribute to risk for neuropsychiatric disorders, but also for disproving other pathways that are less likely to be successful thera-
effects occurring in the opposite direction. While neuropsychiatric pies. One example is the longitudinal study SELECT that studied the
disorder patients have been observed to experience effects on sleep, effect of selenium on prostate cancer risk and included 30,000 partici-
it has been difficult to ascertain whether this is a cause or effect of pants, cost $114 million and lasted 7 years. However, no evidence for
their illness. MR has provided an opportunity to utilize the genetic a protective effect of selenium on prostate cancer was found.65 A
component of neuropsychiatric disorders, sleep and circadian rhythm subsequent MR analysis replicated these findings and was signifi-
to infer causal relationships. Currently, the sleep phenotype with the cantly faster and less expensive.65 Similarly in psychiatric research,
most power is chronotype. The genetic factor of chronotype allows MR analysis may not only be cost-effective but also the most viable
MR analysis to be performed using this phenotype however given that option for examining causal links. It is important to remain cautious of
chronotype is self-reported, it is not an ideal assessment of circadian results from MR. MR replaces an unverifiable set of assumptions
rhythm alterations and further studies may integrate data on measure- (no confounding) with another variable set of
62
ments such as dim-light melatonin onset, which give a more accu- assumptions (no pleiotropy, independence), which may be slightly
rate representation of the impact of sleep and circadian rhythm more reasonable than the observational estimate in certain settings.
alterations on neuropsychiatric disorder risk.61 While MR studies may provide evidence for an association or no asso-
Some areas of concern in the included MR analyses include the ciation, a randomized trial might still be needed to confirm this.
decision in some published studies to lower the p-value threshold for In summary, there is evidence consistent with a causal role for
IVs, which could potentially void the relevance assumption. A p-value neuropsychiatric disorders (ADHD, BD, MDD and SZ) in sleep-related
threshold from GWAS of <5  108 should typically be employed at a phenotypes and evidence that sleep-related phenotypes are associ-
minimum and F-statistics should also be employed to assess the ated with increased risk for neuropsychiatric disorders. These MR
strength of the instrument. However, with larger GWAS and greater results show a bidirectional relationship between neuropsychiatric dis-
utilization of MR, further evidence for the role of circadian rhythm in orders and sleep-related phenotypes and identify potential risk factors
neuropsychiatric disorders is expected. Additionally, in some cases, for follow-up studies.
CRINION ET AL. 11 of 13

ACKNOWLEDGMENTS 11. Liu C, Chung M. Genetics and epigenetics of circadian rhythms and
Many thanks to Dr. Cathy Wyse and Dr. Laura Fahey (Maynooth Uni- their potential roles in neuropsychiatric disorders. Neurosci Bull. 2015;
31(1):141-159. doi:10.1007/s12264-014-1495-3
versity) and to Dr. John Ferguson (University of Galway) for your
12. Roenneberg T. What is chronotype? Sleep Biol Rhythms. 2012;10:75-
helpful comments on this manuscript. Open access funding provided 76. doi:10.1111/j.1479-8425.2012.00541.x
by IReL. 13. Alachkar A, Lee J, Asthana K, et al. The hidden link between circadian
entropy and mental health disorders. Transl Psychiatry. 2022;12(1):
281. doi:10.1038/s41398-022-02028-3
FUND ING INFORMATION
14. Gold AK, Sylvia LG. The role of sleep in bipolar disorder. Nat Sci Sleep.
This research was funded by Science Foundation Ireland (SFI) through 2016;8:207-214. doi:10.2147/NSS.S85754
the SFI Centre for Research Training in Genomics Data Science under 15. Pagani L, St Clair PA, Teshiba TM, et al. Genetic contributions to cir-
grant number 18/CRT/6214 (SC). This publication has emanated from cadian activity rhythm and sleep pattern phenotypes in pedigrees
segregating for severe bipolar disorder. Proc Natl Acad Sci U S A.
research supported in part by grants from SFI under grant number
2016;113(6):E754-E761. doi:10.1073/pnas.1513525113
15/SIRG/3324 and from the European Research Council (ERC) under 16. Lee A, Myung SK, Cho JJ, Jung YJ, Yoon JL, Kim MY. Night shift work
the European Union's Horizon 2020 research and innovation program and risk of depression: meta-analysis of observational studies.
(grant agreement No 950010) (L.M.L.). J Korean Med Sci. 2017;32(7):1091-1096. doi:10.3346/jkms.2017.32.
7.1091
17. Ashton A, Jagannath A. Disrupted sleep and circadian rhythms in
DATA AVAI LAB ILITY S TATEMENT
schizophrenia and their interaction with dopamine signaling. Front
Data sharing not applicable to this article as no datasets were gener- Neurosci. 2020;14:636. Published 2020 Jun 23. doi:10.3389/fnins.
ated or analysed during the current study. 2020.00636
18. Lunsford-Avery JR, Krystal AD, Kollins SH. Sleep disturbances in ado-
lescents with ADHD: a systematic review and framework for future
E TH I CS S T A TE M E N T
research. Clin Psychol Rev. 2016;50:159-174. doi:10.1016/j.cpr.2016.
Data were directly downloaded from published studies and no addi- 10.004
tional ethics approval was needed. Each study is referenced and 19. Geoffray MM, Nicolas A, Speranza M, Georgieff N. Are circadian
details on ethics approval are available in each original study. rhythms new pathways to understand autism spectrum disorder?
J Physiol Paris. 2016;110(4 Pt B):434-438. doi:10.1016/j.jphysparis.
2017.06.002
RE FE R ENC E S 20. Jagannath A, Peirson SN, Foster RG. Sleep and circadian rhythm dis-
1. Pitkanen MLS, Stevens T, Kopelman MD. Neuropsychiatric disorders. ruption in neuropsychiatric illness. Curr Opin Neurobiol. 2013;23(5):
In: Warrell DA, Cox TM, Firth JD, eds. Oxford Textbook of Medicine. 888-894. doi:10.1016/j.conb.2013.03.008
5th ed. Oxford Textbooks; 2010. 10.1093/med/9780199204854. 21. Hansen HG, Starzer M, Nilsson SF, Hjorthøj C, Albert N,
003.2604 Nordentoft M. Clinical recovery and long-term association of special-
2. Cross-Disorder Group of the Psychiatric Genomics Consortium. Iden- ized early intervention services vs treatment as usual among individ-
tification of risk loci with shared effects on five major psychiatric dis- uals with first-episode schizophrenia spectrum disorder: 20-year
orders: a genome-wide analysis. Lancet. 2013;381(9875):1360. doi: follow-up of the OPUS trial. JAMA Psychiatry. 2023;80(4):371-379.
10.1016/S0140-6736(12)62129-1 doi:10.1001/jamapsychiatry.2022.5164
3. First MB. Diagnostic and statistical manual of mental disorders, 5th 22. Lago SG, Tomasik J, van Rees GF, et al. Drug discovery for psychiatric
edition, and clinical utility. J Nerv Ment Dis. 2013;201(9):727-729. doi: disorders using high-content single-cell screening of signaling net-
10.1097/NMD.0b013e3182a2168a work responses ex vivo. Sci Adv. 2019;5(5):eaau9093. doi:10.1126/
4. Doherty JL, Owen MJ. Genomic insights into the overlap between sciadv.aau9093
psychiatric disorders: implications for research and clinical practice. 23. Taber KH, Hurley RA, Yudofsky SC. Diagnosis and treatment of neu-
Genome Med. 2014;6(4):29. ropsychiatric disorders. Annu Rev Med. 2010;61:121-133. doi:10.
5. Jawabri KH, Raja A. Physiology, sleep patterns. StatPearls. StatPearls 1146/annurev.med.051408.105018
Publishing; 2022. 24. Baethge C, Goldbeck-Wood S, Mertens S. SANRA—a scale for the
6. Weidenauer C, Vollmer C, Scheiter K, Randler C. Weak associations quality assessment of narrative review articles. Res Integr Peer Rev.
of Morningness-Eveningness and stability with skin temperature 2019;4:5. doi:10.1186/s41073-019-0064-8
and cortisol levels. J Circadian Rhythms. 2019;17:8. doi:10.5334/ 25. Jansen PR, Watanabe K, Stringer S, et al. Genome-wide analysis of
jcr.182 insomnia in 1,331,010 individuals identifies new risk loci and func-
7. Jones SE, Lane JM, Wood AR, et al. Genome-wide association ana- tional pathways. Nat Genet. 2019;51(3):394-403. doi:10.1038/
lyses of chronotype in 697,828 individuals provides insights into cir- s41588-018-0333-3
cadian rhythms. Nat Commun. 2019;10(1):343. doi:10.1038/s41467- 26. Demontis D, Walters RK, Martin J, et al. Discovery of the first
018-08259-7 genome‐wide significant risk loci for attention deficit/hyperactivity
8. Jones SE, van Hees VT, Mazzotti DR, et al. Genetic studies of disorder. Nat Genet. 2019;51(1):63‐75. doi:10.1038/s41588‐018‐
accelerometer-based sleep measures yield new insights into human 0269‐7
sleep behaviour. Nat Commun. 2019;10(1):1585. doi:10.1038/ 27. Grove J, Ripke S, Als TD, et al. Identification of common genetic risk
s41467-019-09576-1 variants for autism spectrum disorder. Nat Genet. 2019;51(3):431‐
9. Dibner C, Schibler U. Circadian timing of metabolism in animal models 444. doi:10.1038/s41588‐019‐0344‐8
and humans. J Intern Med. 2015;277(5):513-527. doi:10.1111/joim. 28. Mullins N, Forstner AJ, O'Connell KS, et al. Genome-wide association
12347 study of more than 40,000 bipolar disorder cases provides new
10. Baron KG, Reid KJ. Circadian misalignment and health. Int Rev Psychi- insights into the underlying biology. Nat Genet. 2021;53(6):817-829.
atry. 2014;26(2):139-154. doi:10.3109/09540261.2014.911149 doi:10.1038/s41588-021-00857-4
12 of 13 CRINION ET AL.

29. Watanabe K, Jansen PR, Savage JE, et al. Genome‐wide meta‐analysis 48. Bowden J, Del Greco MF, Minelli C, Davey Smith G, Sheehan N,
of insomnia prioritizes genes associated with metabolic and psychiat- Thompson J. A framework for the investigation of pleiotropy in two-
ric pathways. Nat Genet. 2022;54(8):1125‐1132. doi:10.1038/ sample summary data Mendelian randomization. Stat Med. 2017;
s41588‐022‐01124‐w 36(11):1783-1802. doi:10.1002/sim.7221
30. Howard DM, Adams MJ, Clarke TK, et al. Genome‐wide meta‐analy- 49. Carpena MX, Bonilla C, Matijasevich A, et al. Sleep-related traits and
sis of depression identifies 102 independent variants and highlights attention-deficit/hyperactivity disorder comorbidity: shared genetic
the importance of the prefrontal brain regions. Nat Neurosci. 2019;22 risk factors, molecular mechanisms, and causal effects. World J Biol
(3):343‐352. doi:10.1038/s41593‐018‐0326‐7 Psychiatry. 2021;22(10):778-791.
31. Trubetskoy V, Pardiñas AF, Qi T, et al. Mapping genomic loci impli- 50. Sun X, Liu B, Liu S, et al. Sleep disturbance and psychiatric disorders:
cates genes and synaptic biology in schizophrenia. Nature. 2022;604 a bidirectional Mendelian randomisation study. Epidemiol Psychiatr
(7906):502‐508. doi:10.1038/s41586‐022‐04434‐5 Sci. 2022;31:e26. doi:10.1017/S2045796021000810
32. Dashti HS, Jones SE, Wood AR, et al. Genome-wide association study 51. Choi KW, Stein MB, Nishimi KM, et al. An exposure-wide and Men-
identifies genetic loci for self-reported habitual sleep duration sup- delian randomization approach to identifying modifiable factors for
ported by accelerometer-derived estimates. Nat Commun. 2019; the prevention of depression. Am J Psychiatry. 2020;177(10):944-
10(1):1100. doi:10.1038/s41467-019-08917-4 954. doi:10.1176/appi.ajp.2020.19111158
33. Walker WH 2nd, Walton JC, DeVries AC, Nelson RJ. Circadian 52. Cai L, Bao Y, Fu X, et al. Causal links between major depressive disor-
rhythm disruption and mental health. Transl Psychiatry. 2020;10(1): der and insomnia: a Mendelian randomisation study. Gene. 2021;768:
28. doi:10.1038/s41398-020-0694-0 145271. doi:10.1016/j.gene.2020.145271
34. Roybal K, Theobold D, Graham A, et al. Mania-like behaviour induced 53. Chen G, Xie J, Liu W, et al. Association between depression and sleep
by disruption of CLOCK. Proc Natl Acad Sci U S A. 2007;104(15): apnoea: a Mendelian randomisation study. ERJ Open Res. 2021;8(1):
6406-6411. doi:10.1073/pnas.0609625104 00394. doi:10.1183/23120541.00394-2021
35. Schuch JB, Genro JP, Bastos CR, Ghisleni G, Tovo-Rodrigues L. The 54. Mulugeta A, Zhou A, King C, Hyppönen E. Association between major
role of CLOCK gene in psychiatric disorders: evidence from human depressive disorder and multiple disease outcomes: a phenome-wide
and animal research. Am J Med Genet B Neuropsychiatr Genet. 2018; Mendelian randomisation study in the UK Biobank. Mol Psychiatry.
177(2):181-198. doi:10.1002/ajmg.b.32599 2020;25(7):1469-1476. doi:10.1038/s41380-019-0486-1
36. Mogavero F, Jager A, Glennon JC. Clock genes, ADHD and aggres- 55. Baranova A, Cao H, Zhang F. Shared genetic liability and causal
sion. Neurosci Biobehav Rev. 2018;91:51-68. doi:10.1016/j.neubiorev. effects between major depressive disorder and insomnia. Hum Mol
2016.11.002 Genet. 2022;31(8):1336-1345. doi:10.1093/hmg/ddab328
37. Kelly MP. A role for phosphodiesterase 11A (PDE11A) in the forma- 56. Gao X, Meng LX, Ma KL, et al. The bidirectional causal relationships
tion of social memories and the stabilization of mood. Adv Neurobiol. of insomnia with five major psychiatric disorders: A Mendelian ran-
2017;17:201-230. doi:10.1007/978-3-319-58811-7_8 domization study. Eur Psychiatry. 2019;60:79-85. doi:10.1016/j.
38. Watanabe K, Stringer S, Frei O, et al. A global overview of pleiotropy eurpsy.2019.05.004
and genetic architecture in complex traits. Nat Genet. 2019;51(9): 57. Nassan M, Daghlas I, Winkelman JW, Dashti HS, International Suicide
1339-1348. doi:10.1038/s41588-019-0481-0 Genetics Consortium, Saxena R. Genetic evidence for a potential
39. Smith GD, Ebrahim S. ‘Mendelian randomization’: can genetic epide- causal relationship between insomnia symptoms and suicidal behav-
miology contribute to understanding environmental determinants of ior: a Mendelian randomization study. Neuropsychopharmacology.
disease? Int J Epidemiol. 2003;32(1):1-22. doi:10.1093/ije/dyg070 2022;47(9):1672-1679. doi:10.1038/s41386-022-01319-z
40. Ference BA, Holmes MV, Smith GD. Using Mendelian randomization 58. Zhou F, Guo Y, Wang Z, Liu S, Xu H. Assessing the causal associations
to improve the design of randomized trials. Cold Spring Harb Perspect of insomnia with depressive symptoms and subjective well-being: a
Med. 2021;11(7):a040980. doi:10.1101/cshperspect.a040980 bidirectional Mendelian randomization study. Sleep Med. 2021;87:85-
41. Burgess S, Davey Smith G, Davies NM, et al. Guidelines for perform- 91. doi:10.1016/j.sleep.2021.08.025
ing Mendelian randomization investigations. Wellcome Open Res. 59. O'Loughlin J, Casanova F, Jones SE, et al. Using Mendelian randomi-
2020;4:186. doi:10.12688/wellcomeopenres.15555.2 sation methods to understand whether diurnal preference is causally
42. Burgess S, Thompson SG, CRP CHD Genetics Collaboration. Avoiding related to mental health. Mol Psychiatry. 2021;26(11):6305-6316. doi:
bias from weak instruments in Mendelian randomization studies. Int J 10.1038/s41380-021-01157-3
Epidemiol. 2011;40(3):755-764. doi:10.1093/ije/dyr036 60. Daghlas I, Lane JM, Saxena R, Vetter C. Genetically proxied diurnal pref-
43. Davey Smith G, Hemani G. Mendelian randomization: genetic anchors erence, sleep timing, and risk of major depressive disorder. JAMA Psychi-
for causal inference in epidemiological studies. Hum Mol Genet. 2014; atry. 2021;78(8):903-910. doi:10.1001/jamapsychiatry.2021.0959
23(R1):R89-R98. doi:10.1093/hmg/ddu328 61. Burgess S, Foley CN, Allara E, Staley JR, Howson JMM. A robust and
44. Davies NM, Holmes MV, Davey SG. Reading Mendelian randomisa- efficient method for Mendelian randomization with hundreds of
tion studies: a guide, glossary, and checklist for clinicians. BMJ. 2018; genetic variants. Nat Commun. 2020;11(1):376. doi:10.1038/s41467-
362:k601. doi:10.1136/bmj.k601 019-14156
45. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with 62. Kennaway DJ. The dim light melatonin onset across ages, methodolo-
invalid instruments: effect estimation and bias detection through gies, and sex and its relationship with morningness/eveningness.
egger regression. Int J Epidemiol. 2015;44(2):512-525. doi:10.1093/ Sleep. 2023;46(5):zsad033. doi:10.1093/sleep/zsad033
ije/dyv080 63. Porcu E, Rüeger S, Lepik K, et al. Mendelian randomization integrating
46. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent esti- GWAS and eQTL data reveals genetic determinants of complex and
mation in Mendelian randomization with some invalid instruments clinical traits. Nat Commun. 2019;10(1):3300. doi:10.1038/s41467-
using a weighted median estimator. Genet Epidemiol. 2016;40(4):304- 019-10936-0
314. doi:10.1002/gepi.21965 64. Lippman SM, Klein EA, Goodman PJ, et al. Effect of selenium and
47. Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary vitamin E on risk of prostate cancer and other cancers: the selenium
data Mendelian randomization via the zero modal pleiotropy assump- and vitamin E cancer prevention trial (SELECT). JAMA. 2009;301(1):
tion. Int J Epidemiol. 2017;46(6):1985-1998. doi:10.1093/ije/dyx102 39-51. doi:10.1001/jama.2008.864
CRINION ET AL. 13 of 13

65. Yarmolinsky J, Bonilla C, Haycock PC, et al. Circulating selenium and

prostate cancer risk: a Mendelian randomization analysis. J Natl Can- How to cite this article: Crinion S, Morris DW, Lopez LM.
cer Inst. 2018;110(9):1035-1038. doi:10.1093/jnci/djy081
Neuropsychiatric disorders, chronotype and sleep: A narrative
review of GWAS findings and the application of Mendelian
SUPPORTING INFORMATION randomization to investigate causal relationships. Genes, Brain
Additional supporting information can be found online in the Support- and Behavior. 2024;23(1):e12885. doi:10.1111/gbb.12885
ing Information section at the end of this article.