Relationship between sleep characteristics and aortic aneurysm/dissection: a two-sample Mendelian randomisation study

Abstract

Objective

The causal relationship between sleep characteristics and aortic aneurysm and dissection (AAD) is little known.

Methods

In this two-sample Mendelian randomisation (MR) study, we selected seven sleep-related traits (sleep duration, getting up in the morning, chronotype, nap during day, insomnia, snoring, and narcolepsy) from published genome-wide association study (GWAS)-related genetic variants as instrumental variables. Causality was assessed by two-sample MR analysis using inverse-variance weighting (IVW), MR-Egger regression, weighted median, weighted mode and simple model. Horizontal pleiotropy was tested using MR-Egger regression and MR-polytropic residuals and outliers, and heterogeneity was calculated by Cochran’s Q test.

Results

There was no evidence of causality among sleep duration (IVW: OR=0.759, 95% CI: 0.489 to 1.177, p=0.218), getting up in the morning (IVW: OR=1.148, 95% CI: 0.768 to 1.716, p=0.502), chronotype (IVW: OR=0.960, 95% CI: 0.796 to 1.158, p=0.670), nap during day (IVW: OR=1.248, 95% CI: 0.771 to 2.020, p=0.367), sleeplessness/insomnia (IVW: OR=1.280, 95% CI: 0.678 to 2.414, p=0.447), snoring (IVW: OR=0.963, 95% CI: 0.770 to 1.203, p=0.738), narcolepsy (IVW: OR=1.025, 95% CI: 0.367 to 2.863, p=0.962) and aortic aneurysm. Moreover, there was no evidence to suggest a causal relationship among sleep traits and abdominal aneurysm, thoracic aneurysm and aortic dissection. Sensitivity analyses, including leave-one-out, horizontal pleiotropy and heterogeneity tests, indicated that our results were robust and reliable.

Conclusions

Overall, our study found no genetic evidence of a causal relationship between sleep characteristics and AAD. Large-scale randomised controlled trial experiments are also needed to further verify the causal relationship between sleep and AAD.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study used a two-sample Mendelian randomisation (MR) design, which helps to minimise confounding and reverse causality, providing a more reliable estimation of causal effects.

• We employed multiple MR analytical methods and conducted comprehensive sensitivity analyses to ensure the robustness and reliability of our findings.

• The instrumental variables were derived from large-scale genome-wide association studies (GWAS), enhancing the statistical power of our causal inferences.

• Despite the rigorous methodology, the generalisability of our findings may be limited to individuals of European ancestry, as the GWAS summary statistics were predominantly based on European populations.

• The F-statistic for narcolepsy is <10, indicating the presence of weak instrument variable bias, which may affect the reliability of the result that narcolepsy is not causally related to aortic aneurysm and dissection.

Introduction

Aortic aneurysm and dissection (AAD) is the second most common aortic disease after atherosclerosis, and it ranks 19th in causes of death in the USA, with approximately 43 000–47 000 deaths annually.¹ Aortic aneurysm (AA) is characterised by the dilation of the aortic wall, whereas aortic dissection (AD) typically results from an intimal tear, allowing blood to enter the medial layer and split it apart. Depending on the location of the AA, it can be classified as thoracic AA (TAA) and abdominal AA (AAA). AA is mostly asymptomatic and insidious, but once progressed to AD or aortic rupture, they cause sudden death or severe disability, resulting in incalculable human and economic losses.¹ For acute type A aortic dissection, the mortality rate increases by 1–2% for every hour after the symptoms onset.² The possible pathogenesis of AAD includes Vascular smooth muscle cell (VSMC) phenotypic switching, apoptosis, autophagy, oxidative stress and release of various cytokines, such as IL-1 and IL-6 by immune cells,³ but the specific aetiology is more complex and difficult to predict. Therefore, understanding the risk factors of AAD to prevent the incidence of AAD is of great importance.

Sleep deprivation is becoming increasingly prevalent, and altered sleep characteristics such as sleep quality, sleep duration and sleep habits may pose a significant threat to human health.⁴ Previous evidence has shown that sleep is strongly associated with morbidity and mortality of cardiovascular diseases (CVDs) such as hypertension, coronary heart disease (CHD) and stroke.⁴,⁷ People with poor or average sleep have a 39% higher mortality rate from CVDs than normal individuals.⁶ In addition, a prospective cohort study in a middle-aged and elderly population found that sleep characteristics such as sleep duration, insomnia, snoring, daytime sleepiness and chronotype also affect the incidence of CVDs.⁸ Obstructive sleep apnoea syndrome (OSAS), characterised by partial and complete upper airway obstruction, has been identified to have a high prevalence in hypertension, heart failure and many other CVDs, with prevalence rates as high as 40–80%,⁹ while these CVDs may subsequently lead to aortic rupture and dissection. In mice, OSAS promotes the onset and progression of aortic dissection (AD) disease through the ROS-HIF-1α-MMPs signalling axis.¹⁰ Results from several cohort studies have shown that OSAS is associated with the prevalence and progression of AAD.^{11 12} However, no study has reported whether there is a causal relationship between sleep characteristics and AAD.

Mendelian randomisation (MR) studies are a statistical method that assumes genetic variation of single nucleotide polymorphism (SNP) reflecting exposure as valid instrumental variables (IVs) for exploring whether there is a causal relationship between exposure and outcome in epidemiological studies.¹³ When clinical randomised controlled trials (RCTs) are not conducted or feasible due to the presence of confounding factors, MR could effectively reduce the effects of confounding bias and reverse causality by using genetic variation with low bias. Genome-wide association study (GWAS) data incorporate genetic variation and are the primary source of data for MR studies to assess the aetiology of diseases.¹⁴ Inverse-variance weighting (IVW) was selected as the primary MR method because it offers the greatest statistical efficiency when either (1) all SNPs are valid instruments or (2) any violations of the exclusion-restriction assumption are directionally balanced.^{15 16} In this study, we gathered published data on sleep-related traits and AAD from large genetic studies, then compiled summary statistics using GWAS from the Pan-UK BioBank and FinnGen R9 and performed a two-sample MR analysis mainly using the IVW method to assess the causal relationship between sleep traits and AAD.

Methods

Study design

A two-sample MR design was applied to assess the potential causal relationship between seven sleep characteristics (sleep duration, getting up in morning, chronotype, nap during day, sleeplessness/insomnia, snoring and narcolepsy) and AAD (figure 1). Sleep characteristics were defined as the exposure, and AAD including AA, AAA, TAA and AD served as the outcome.

Figure 1. Design of this two-sample Mendelian randomisation study. AA, aortic aneurysm; AAA, abdominal AA; AD, aortic dissection; GWAS, genome-wide association study; MAF, minor allele frequency; MR, Mendelian randomisation; LD, linkage-disequilibrium; PRESSSO, polytropic residuals and outliers; SNP, single nucleotide polymorphism; TAA, thoracic AA.

There are three main assumptions in IVs for sleep characteristics as below: (1) relevance assumption: genetic IVs should be strongly correlated with sleep characteristics, (2) exclusion restriction: genetic IVs should be correlated with the AAD only through sleep characteristics, (3) independence assumption: genetic IVs should not be correlated with any measured or unmeasured confounders.¹³

Relevance assumption was established by selecting SNPs with genome-wide significance (p< 5 ×10⁻⁸) and calculating F-statistics. Exclusion restriction was established by leave-one-out test, MR Egger tests and MR polytropic residuals and outliers (MR-PRESSO) for testing and correcting horizontal pleiotropy. The independence assumption was established by partitioning linkage-disequilibrium (LD) pruning.

Data sources

We extracted genetic association data for sleep duration, getting up in the morning, chronotype, nap during the day, sleeplessness/insomnia, snoring and narcolepsy from GWAS published by UK BioBank covering 418 009, 419 431, 375 390, 420 148, 420 013, 391 531 and 418 874 individuals of European ancestry (online supplemental table S1), respectively.

The GWAS data on AA, AAA, TAA and AD were retrieved from FinnGen R9, which consisted of 7395/3510/3548/881 cases (defined by the International Classification of Diseases (ICD)-Eighth, Ninth and Tenth Revisions) and 349 539 controls of European ancestry, respectively. To reduce group stratification differences, all data were from European cohorts.

The data sources for the above seven sleep-related characteristics and 4 subtypes of AAD have been summarised (online supplemental table S1), and details of the definition and measurement of each sleep-related characteristic are shown in table 1.

Table 1. Description of the sleep traits in the GWAS included in MR analyses.

Trait	Definition	Type of trait	N
Sleep duration	‘About how many hours sleep do you get in every 24 hours? (please include naps)?’ with hour increments.	Continuous	418 009
Getting up in morning	‘On an average day, how easy do you find getting up in the morning?’ with one of the five possible answers: (1) ‘very easy’; (2) ‘easy’; (3) ‘common’; (4) ‘difficult’; (5) ‘very difficult’.	Ordered categorical	419 431
Morning/evening person (chronotype)	‘Do you consider yourself to be?’ with one of the six possible answers: (1) ‘Definitely a ‘morning’ person’; (2) ‘More a ‘morning’ than ‘evening’ person’; (3) ‘More an ‘evening’ than a ‘morning’ person’; (4) ‘Definitely an ‘evening’ person’; (5) ‘Do not know’ and (6) ‘Prefer not to answer’.	Ordered categorical	375 390
Nap during day	‘Do you have a nap during the day?’ with responses of ‘Yes’, ‘No’.	Binary	420 148
Sleeplessness/insomnia	‘Do you have trouble falling asleep at night or do you wake up in the middle of the night?’ with responses of ‘Yes’, ‘No’.	Binary	420 013
Snoring	‘Does your partner or a close relative or friend complain about your snoring?’ with responses of ‘Yes’, ‘No’.	Binary	391 531
Daytime dozing/sleeping (narcolepsy)	‘How likely are you to doze off or fall asleep during the daytime when you don't mean to?’ (1) ‘never’; (2) ‘sometimes’; (3) ‘often’ or (4) ‘all of the time’.	Ordered categorical	418 874

GWAS, genome-wide association study; MR, Mendelian randomisation.

Selection of instrumental variables

Steps for selecting genetic variants: (1) SNPs with a p value<5×10^-8 associated with each sleep phenotype (exposure) were selected as IVs; (2) the minor allele frequency (MAF) threshold of significant SNPs was 0.01; (3) exclude LD among SNPs (R²<0.001, clumping distance=10 000 kb); (4) remove SNPs of the palindromic structure; (5) SNP proxy: if SNPs are not available in the outcome, proxy SNP with R²>0.8 will selected to replace through LDlink (https://ldlink.nih.gov/?tab=ldproxy); (6) the effects of SNPs on the exposure and the outcome need to correspond to the same allele and (7) compute the F-statistic to eliminate bias in the results due to weak instrumental variables (F<10). The formula is as follows: the F-statistic for a single SNP is beta²/SE², and the total F-statistic is calculated as (N−k−1)/k×R²/(1−R²). R² is calculated as 2×MAF×(1−MAF)×beta².

Statistical analysis

In this two-sample MR study, IVW, MR-Egger, weighted median, weighted mode and simple mode were applied to determine whether there was a causal relationship between sleep characteristics and AAD risk. IVW, which assumes that all genetic variants are valid instruments or that any violations of the exclusion restriction are balanced, was used as the primary method for MR analysis on the premise that there was no pleiotropy of SNPs. MR-Egger regression was conducted to detect horizontal pleiotropy, and the result was denoted by the intercept term. Weighted median was used to exclude invalid SNPs, and a weighted mode was used to test the ability of causal effect. Subsequently, we adopted Cochran’s Q statistic to examine heterogeneity. In addition, the application of MR-PRESSO for testing and correcting horizontal pleiotropy. When results from alternative approaches such as MR-Egger regression or the weighted-median estimator differed from the IVW estimate, we interpreted the findings cautiously. Specifically, if the MR-Egger intercept indicated directional pleiotropy, or if the weighted-median effect size diverged materially from the IVW estimate, we placed greater weight on results that were consistent across multiple methods and carefully examined heterogeneity and pleiotropy statistics before drawing conclusions. Ultimately, leave-one-out analysis is used to detect sensitivity to assess the stability of results.

Two-sample MR analyses were performed using the ‘TwoSampleMR’ package. The ‘MRPRESSO’ package was used to conduct the MR-PRESSO test. All statistical analyses were performed in R (V.4.2.2). P value <0.05 was considered statistically significant.

Results

The data sources used were shown in online supplemental table S2. We finally selected 56 SNPs for sleep duration, 76 for getting up in the morning, 144 for chronotype, 82 for nap during the day, 45 for sleeplessness/insomnia, 32 for snoring and 33 for narcolepsy as IVs. Details of the IVs are presented in online supplemental table S2. The results of heterogeneity analysis, horizontal pleiotropy test, leave-one-out analysis and single SNP analysis are presented in online supplemental tables S3–S7. The visualisations of the MR analysis are given in onlinesupplemental figure S1S7.

Sleep duration and AAD

According to the results of MR analysis, sleep duration had no genetic evidence of causal relationship with AA (IVW: OR=0.759, 95% CI: 0.489 to 1.177, p=0.218), AAA (IVW: OR=0.628, 95% CI: 0.319 to 1.237, p=0.178), TAA (IVW: OR=0.793, 95% CI: 0.413 to 1.521, p=0.485) and AD (IVW: OR=0.711, 95% CI=0.219–2.312, p=0.570) (figure 2 and online supplemental figure S1).

Figure 2. Mendelian randomisation estimates of the association between sleep traits and aortic aneurysm/dissection.

Heterogeneity analysis indicated that no heterogeneity was found among the SNPs in AA, AAA, TAA and AD (onlinesupplemental table S3 figure S1). MR-Egger and MR-PRESSO were used to explore the horizontal pleiotropy, and the results showed that there is no horizontal pleiotropy in AA, TAA and AD (online supplemental table S4). However, there was potential horizontal pleiotropy in AAA (Egger intercept=−0.043, p=0.007, online supplemental table S4), but MR-PRESSO showed no horizontal pleiotropy (global p=0.086) with no significant outlier. Leave-one-out analysis and single SNP analysis showed that the MR analysis results were robust and reliable (onlinesupplemental tables S5, S6 figure S1). The F-statistic (24.824) of IVs was >10, indicating that there is no weak instrumental variable bias.

Getting up in the morning and AAD

According to the MR analysis, getting up in the morning is not causally related to the AA (IVW: OR=1.148, 95% CI: 0.768 to 1.716, p=0.502), AAA (IVW: OR=1.175, 95% CI: 0.643 to 2.148, p=0.600), TAA (IVW: OR=1.080, 95% CI: 0.620 to 1.880, p=0.787) and AD (IVW: OR=0.632, 95% CI=0.216 to 1.848, p=0.402) (figure 2, online supplemental figure S2).

No heterogeneity was found between SNPs in AA, AAA, TAA and AD (onlinesupplemental table S3 figure S2). Moreover, no horizontal pleiotropy was found in AA, AAA, TAA and AD (online supplemental table S4). The leave-one-out analysis and single SNP analysis indicated that the results were robust and reliable (onlinesupplemental table S5, S6 figure S2). There was no weak instrumental variable bias (F=22.562).

Morning/evening person (chronotype) and AAD

Results showed that chronotype has no genetic evidence of causal relationship with AA (IVW: OR=0.960, 95% CI: 0.796 to 1.158, p=0.670), AAA (IVW: OR=1.106, 95% CI: 0.842 to 1.451, p=0.469), TAA (IVW: OR=0.827, 95% CI: 0.642 to 1.064, p=0.140) and AD (IVW: OR=0.976, 95% CI=0.589 to 1.616, p=0.924) (figure 2, online supplemental figure S3).

No heterogeneity was found among SNPs in AA, TAA and AD. The existence of heterogeneity was found in AAA (Q=177.337, p=0.024, onlinesupplemental table S3 figure S3). There was no horizontal pleiotropy among chronotype and AA, AAA, TAA and AD by MR-Egger regression (online supplemental file 1). But horizontal pleiotropy was found in AAA by MR-PRESSO (global p=0.025) with no significant outlier. Leave-one-out analysis and single SNP analysis results indicate that the MR analysis results are robust and reliable (onlinesupplemental tables S5,S6 figure S3). No weak instrumental variable bias exists (F=49.871).

Nap during day and AAD

The absence of a causal relationship among nap during day sleep and AA (IVW: OR=1.248, 95% CI: 0.771 to 2.020, p=0.367), AAA (IVW: OR=1.548, 95% CI: 0.757 to 3.167, p=0.231), TAA (IVW: OR=1.002, 95% CI: 0.541 to 1.860, p=0.993) and AD (IVW: OR=0.718, 95% CI=0.212 to 2.437, p=0.596) (figure 2 and online supplemental figure S4).

The presence of heterozygosity was not found among any of the SNPs of nap during day and AA, TAA and AD, while heterozygosity exists in AAA (Q=106.425, p=0.026, onlinesupplemental table S3 figure S4). We found no horizontal pleiotropy or significant outlier among nap during day and AA, AAA, TAA and AD by using MR-Egger regression (online supplemental table S4) and MR-PRESSO method. The results of sensitivity analyses showed that this MR analysis was robust and reliable (onlinesupplemental tables S5, S6 figure S4). There is no weak instrumental variable bias (F=16.301).

Sleeplessness/insomnia and AAD

Insomnia was not causally correlated with AA (IVW: OR=1.280, 95% CI: 0.678 to 2.414, p=0.447), AAA (IVW: OR=1.229, 95% CI: 0.464 to 3.256, p=0.678), TAA (IVW: OR=1.064, 95% CI: 0.444 to 2.550, p=0.889) and AD (IVW: OR=1.309, 95% CI=0.269 to 6.368, p=0.738) according to the two-sample MR analysis (figure 2, online supplemental figure S5).

When testing for heterogeneity using the Cochrane Q statistic, no heterogeneity was found in AD, while heterogeneity was found among SNPs in AA (Q=70.473, p=0.005), AAA (Q=79.666, p=0.001) and TAA (Q=66.489, p=0.012) (onlinesupplemental table S3 figure S5). It was found that there was no horizontal pleiotropy between insomnia and AA, AAA, TAA and AD by MR-Egger. However, horizontal pleiotropy was found between insomnia and AA (global p=0.006), AAA (global p<0.001), TAA (global p=0.030) by MR-PRESSO. The results of leave-one-out analysis indicated that the results were robust and reliable (onlinesupplemental tables S5, S6 figure S5). There was no weak instrumental variable bias (F=20.893).

Snoring and AAD

It was found that there was no genetic evidence of causal correlation between snoring and AA (IVW: OR=0.963, 95% CI: 0.770 to 1.203, p=0.738), AAA (IVW: OR=0.897, 95% CI: 0.664 to 1.212, p=0.478), TAA (IVW: OR=0.939, 95% CI: 0.694 to 1.271, p=0.685) and AD (IVW: OR=0.962, 95% CI=0.544 to 1.702, p=0.894) (figure 2, online supplemental figure S6).

No significant heterogeneity was found among SNPs in AA, AAA, TAA and AD (onlinesupplemental table S3 figure S6). No horizontal pleiotropy was found between snoring and AAD by MR-Egger (online supplemental table S4), but found between snoring and AA by MR-PRESSO with no outlier (global p=0.048). The results of this MR analysis were robust and reliable (onlinesupplemental tables S5, S6 figure S6). No weak instrumental variable bias exists (F=205.352).

Daytime dozing/sleeping (narcolepsy) and AAD

Narcolepsy was not causally associated with AA (IVW: OR=1.025, 95% CI: 0.367 to 2.863, p=0.962), AAA (IVW: OR=0.674, 95% CI: 0.145 to 3.139, p=0.615), TAA (IVW: OR=0.603, 95% CI: 0.170 to 2.143, p=0.435) and AD (IVW: OR=0.457, 95% CI=0.040 to 5.262, p=0.530) (figure 2, online supplemental figure S7).

No heterogeneity was present in AA, TAA and AD, while heterogeneity was present in AAA (Q=48.433, p=0.024) (onlinesupplemental table 3 figure S7). There was no horizontal pleiotropy between narcolepsy and AA, AAA, TAA and AD by MR-Egger (online supplemental table S4) and MR-PRESSO. The leave-one-out analysis and single SNP analysis indicated that the results of this MR analysis were robust and reliable (onlinesupplemental tables S5, S6 figure S7). The F-statistic (9.926) of IVs was <10, indicating that there was weak instrumental variable bias.

Discussion

Several sleep characteristics, including sleep duration, insomnia with circadian rhythm disruption and prolonged daytime napping, are closely associated with a high incidence and mortality rate of CVDs such as hypertension, CHD and stroke.^{4 7 17 18} Insufficient sleep directly increases the risk of CVDs by altering vascular structure and function, coronary microcirculation and impairing endothelial function through various mechanisms.⁴¹⁹,²¹ Additionally, insufficient sleep indirectly raises the risk of CVDs by disrupting glucose homeostasis, reducing insulin sensitivity, increasing obesity, the incidence of type 2 diabetes and impairing immune function through endocrine pathways.^{22 23}

Previous studies have shown that risk factors for hypertension, CHD and similar conditions are also high-risk factors for AAD, and they are positively correlated with the occurrence and progression of AAD.²⁴,²⁶ Observational meta-analyses and cohort studies have indicated that OSA, a frequently occurring sleep-related breathing disorder, is associated with an increased risk of AAD and is an independent risk factor for TAA expansion.^{11 27} However, another meta-analysis and cohort study results have found no association between sleep-disordered breathing and the risk of AA and ADs.^{28 29} Therefore, the impact of OSA on AAD remains controversial. In addition, the present study also failed to detect genetic evidence of a causal relationship between the seven sleep characteristics and the occurrence and progression of AAD. These results indicated that sleep may not directly influence the pathogenesis of AAD but may promote the development of AAD through its impact on CVDs such as CHD and stroke.

Although AD may occur as a complication of AA, they often arise independently. A nationwide registry study from Australia found that the majority of patients with acute AD had normal aortic diameters prior to onset, suggesting that AD is not merely an advanced stage of AA but a clinically distinct event.³⁰ What’s more, the thoracic and abdominal segments of the aorta arise from different embryological lineages: neural crest cells for the thoracic aorta and paraxial mesoderm for the abdominal aorta, resulting in regional differences in gene expression, extracellular matrix composition and responsiveness to signalling pathways.³¹ Moreover, the density of α-adrenergic receptors varies across segments, influencing vascular tone and reactivity. These structural and molecular differences, coupled with distinct haemodynamic stresses (eg, pulse wave reflections, wall shear stress), contribute to the divergent pathobiology of AAA and TAA, warranting their separate analysis. In addition, the risk factors of TAA and AAA also differ significantly. Risk factors for AAA mainly include smoking, age >60 years, atherosclerosis, hypertension, male sex, first-degree relatives with AAA, dyslipidaemia and obesity.³² Risk factors for TAA primarily include atherosclerosis, smoking, syndromic disorders (such as Marfan syndrome), non-syndromic disorders, bicuspid aortic valve, infectious or non-infectious aortitis and trauma.³² Additionally, single-cell sequencing studies have shown differences in biological functions and pathogenic genes between AAA and TAA.³³ Multiple GWAS have identified partially distinct genetic architectures for AAA and TAA. For example, LDLR and SORT1 are associated with AAA, whereas FBN1, LOX and ACTA2 mutations are more closely related to TAA and AD.³⁴,³⁶ This genetic divergence provides further justification for stratified MR analysis. Therefore, we analysed the effects of sleep characteristics on AA, AAA, TAA and AD, aiming to provide a more accurate interpretation and potential future applications for personalised prevention strategies based on subtype-specific mechanisms. Unfortunately, our results do not support genetic evidence of a causal relationship between sleep characteristics and various types of AAD.

Limitations

This study also has certain limitations: First, there is heterogeneity in some of the sleep characteristics studied (chronotype, nap during the day, sleeplessness/insomnia, narcolepsy), which may be caused by various factors, including but not limited to subtype differences between different genes, gene interactions, statistical variation, randomness, etc. Second, the F-statistic for narcolepsy is <10, indicating the presence of weak instrument variable bias, which may affect the reliability of the result that narcolepsy is not causally related to AAD. Although we verified that all individual SNPs used as instruments for narcolepsy had F-statistics >10 (online supplemental table S2), the limited variance explained may still introduce bias or imprecision. Therefore, the null result for narcolepsy should be interpreted cautiously. Additional GWASs with larger sample sizes are warranted to enhance the statistical power of MR studies for narcolepsy. However, the F-statistic itself has limitations. It mainly measures the strength of the relationship between IVs and the outcome variable, but it does not directly control for the validity of IVs.³⁷ It can only detect linear models and cannot detect non-linear or more complex relationships well. Additionally, it only tests whether the population means under the null hypothesis are equal and does not account for individual variability. Therefore, the F-value may not fully reflect the association between narcolepsy and AAD, despite the results that narcolepsy and AAD are not causally related. Our results still support the absence of genetic evidence of causal relationship between sleep and AAD. Third, all sleep characteristics in this study were subjectively reported by patients, and misclassification cannot be avoided. Fourth, the population studied in this research was limited to European populations. The GWAS summary statistics for both exposures (sleep characteristics) and outcomes (AAD subtypes) in our study were obtained from the UK Biobank and FinnGen, two of the largest and most well-characterised population biobanks in Europe. Both resources are widely accepted as representative large-scale datasets for genetic epidemiology among individuals of European ancestry.^{38 39} These cohorts are frequently used in pan-European genome-wide association meta-analyses and are cited as reference datasets by the GWAS Catalog and European Bioinformatics Institute. Although the UK Biobank predominantly consists of individuals of White British ancestry (~94%), haplotype-based analyses have shown that the cohort captures much of the common genetic diversity across northwestern European populations.⁴⁰ Similarly, FinnGen represents a genetically isolated but homogeneous Finnish population; it offers an important replication context within northern Europe. Therefore, our findings should not be assumed to apply to Mediterranean, Slavic or non-European ancestries without further validation. It is necessary to further verify and supplement other populations and assess the results at multiple levels.

Conclusion

Overall, our study found no genetic evidence of a causal relationship between sleep characteristics and AAD. Large-scale RCT experiments are also needed to further verify the causal relationship between sleep and AAD.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.