Causal Associations between Sleep Traits, Sleep Disorders, and Glioblastoma: A Two-Sample Bidirectional Mendelian Randomization Study

This study applied Mendelian Randomization (MR) analysis to investigate the bidirectional causal association of sleep traits (chronotype, getting up in the morning, midday nap, sleep duration, and sleep episodes) and sleep disorders (insomnia, narcolepsy, sleep apnea, and general sleep disorders) with GBM, based on their potential links with GBM.
Study design and data source
GWAS data for GBM was sourced from genome sequencing data of the Finn cohort. Data regarding sleep traits and sleep disorders were collected from the UKB and GWAS Catalog, with sample sizes ranging from 84,810 to 462,400 (Table S1).
MR analysis
A bidirectional MR approach was used to investigate the causal associations between sleep traits, sleep disorders, and GBM.
Sensitivity analysis
Sensitivity analysis was conducted to detect potential heterogeneity in MR studies. Cochran's Q test was used to assess heterogeneity among IVs, with P > 0.05 indicating low heterogeneity, meaning the estimates among instrumental variables were randomly distributed and had little impact on IVW results. Considering the potential impact of genetic variation on the estimation of association effects, this study used the MR-Egger regression method to explore the presence of horizontal pleiotropy; when the intercept of MR-Egger regression approaches zero or is not statistically significant, it suggests the absence of pleiotropy. Additionally, the MR pleiotropy residual sum and outlier (MR-PRESSO) method was used to detect potential outliers (i.e., SNPs with P < 0.05) and re-estimate causal associations after their removal to correct for horizontal pleiotropy. Leave-one-out analysis was employed to assess the robustness and consistency of the results.
RESULTS
The causal effects of sleep traits and sleep disorders on GBM
First, we assessed the causal effects of sleep traits and sleep disorders on GBM risk. For sleep traits, 152 IVs were selected for chronotype, 73 for getting up in the morning, 84 for a midday nap, 68 for sleep duration, and 20 for sleep episodes. For sleep disorders, 39 IVs were selected for insomnia, 30 for narcolepsy, 16 for sleep apnea, and 27 for general sleep disorders. The F values are shown in Table S2. No associations were found between other sleep traits or any of the sleep disorders and GBM (all P > 0.05) (Figure S1-4).
The causal effect of GBM on sleep traits and sleep disorders
The causal effects of GBM on sleep traits and sleep disorders were assessed. A total of 4 IVs were selected; the F values are shown in Table S2. No association was found when MR Egger and Weighted mode were applied (both P > 0.05), while WM showed statistical significance (OR: 1.0085, 95% CI: 1.0011 - 1.0159, P = 0.024). There was no association between GBM and other sleep traits and sleep disorders (all P > 0.05) (Figure S5-8).

本研究采用孟德尔随机化（MR）分析方法，旨在探究睡眠特征（包括睡眠时相、晨起时间、午后小憩、睡眠时长及睡眠阶段）与睡眠障碍（如失眠、嗜睡、睡眠呼吸暂停及一般睡眠障碍）与胶质母细胞瘤（GBM）之间的双向因果关联，基于其与GBM之间潜在联系的考量。

研究设计与数据来源
胶质母细胞瘤（GBM）的全基因组关联研究（GWAS）数据来源于芬兰队列的基因组测序数据。睡眠特征与睡眠障碍的相关数据收集自英国生物样本库（UKB）和GWAS目录，样本量介于84,810至462,400之间（见表S1）。

孟德尔随机化分析
本研究采用了双向孟德尔随机化方法来探究睡眠特征、睡眠障碍与GBM之间的因果关联。

敏感性分析
为了检测孟德尔随机化研究中潜在的同质性，本研究进行了敏感性分析。使用Cochran Q检验来评估工具变量的异质性，P值大于0.05表明异质性低，意味着工具变量之间的估计是随机分布的，对逆方差加权（IVW）结果的影响甚微。考虑到遗传变异对关联效应估计的潜在影响，本研究采用了MR-Egger回归方法来探究横断面多效性的存在；当MR-Egger回归的截距接近零或无统计学显著性时，表明不存在多效性。此外，本研究还使用了MR多效性残留和异常值（MR-PRESSO）方法来检测潜在的异常值（即P值小于0.05的单核苷酸多态性），并在移除这些异常值后重新估计因果关联以校正横断面多效性。采用留一法分析来评估结果的稳健性和一致性。

结果
睡眠特征与睡眠障碍对GBM的因果效应
首先，我们评估了睡眠特征与睡眠障碍对GBM风险的影响。对于睡眠特征，选择了152个工具变量用于睡眠时相，73个用于晨起时间，84个用于午后小憩，68个用于睡眠时长，20个用于睡眠阶段。对于睡眠障碍，选择了39个工具变量用于失眠，30个用于嗜睡，16个用于睡眠呼吸暂停，27个用于一般睡眠障碍。F值显示在表S2中。未发现其他睡眠特征或任何睡眠障碍与GBM之间存在关联（所有P值均大于0.05）（图S1-4）。

GBM对睡眠特征与睡眠障碍的因果效应
评估了GBM对睡眠特征与睡眠障碍的因果效应。共选择了4个工具变量；F值显示在表S2中。当应用MR-Egger和加权模式时，未发现关联（两者P值均大于0.05），而加权模式显示了统计学显著性（OR：1.0085，95% CI：1.0011 - 1.0159，P = 0.024）。GBM与其他睡眠特征和睡眠障碍之间没有关联（所有P值均大于0.05）（图S5-8）。