Sleep Characteristics and Risk of Stroke and Dementia An Observational and Mendelian Randomization Study
睡眠特征与卒中和痴呆的风险：一项观察性和孟德尔随机化研究
翻译：研究生：张洮铭；科研秘书：刘虹；导师：陈晨
导语：脑卒中以及痴呆是全球严重的长期残疾和死亡的主要原因，也是我院神经内科收治最多的病种，这篇来自1区7.7分的2024年最新文献提示，多种睡眠与卒中和痴呆发生的风险相关，但这些控制心血管危险因素后，关联性减弱。
这是一篇关于睡眠水平和卒中以及痴呆关系调查研究，使用了孟德尔随机化的分析方法。高质量的文献和研究既提示我们选择合适的分析方式的重要性，也给我们科学研究带来了新的思路。通读本文后，给我们研究者带来继续进行深入研究的思路和冲动。身处心血管病专科医院，上述病例和数据不难获得，要的就是十年如一日的科研坚持和随访。
下面我们就来阅读整篇文献。
Abstract
Background and Objectives
Sleep disturbances are implicated as risk factors of both stroke and dementia. However, whether these associations are causal and whether treatment of sleep disorders could reduce stroke and dementia risk remain uncertain. We aimed to evaluate associations and ascertain causal relation-ships between sleep characteristics and stroke/dementia risk and MRI markers of small vessel disease (SVD).
背景与目的
睡眠障碍是脑卒中和痴呆的危险因素.然而，这些关联是否存在因果关系，以及睡眠障碍的治疗是否可以降低中风和痴呆症的风险仍然不确定。本研究旨在评估睡眠特征与卒中/痴呆风险和小血管疾病（SVD）MRI标记物之间的关联并确定因果关系。
Methods
We used data sets from a multicenter population-based study and summary statistics from genome-wide association studies (GWASs) of sleep characteristics and outcomes. We analyzed 502,383 UK Biobank participants with self-reported sleep measurements, including sleep duration, insomnia, chronotype, napping, daytime dozing, and snoring. In observational analyses, the primary outcomes were incident stroke, dementia, and their subtypes, alongside SVD markers. Hazard ratios (HRs) and odds ratios (ORs) were adjusted for age, sex, and ethnicity, and additional vascular risk factors. In Mendelian randomization (MR) analyses, ORs or risk ratios are reported for the association of each genetic score with clinical or MRI end points.
方法
我们使用了多中心人群研究的数据集和睡眠特征和结果的全基因组关联研究（GWAS）的汇总统计数据。分析了502383名英国生物库参与者的自我报告睡眠测量结果，包括睡眠时间、失眠、睡眠时间类型、午睡、白天打瞌睡和打鼾。在观察性分析中，主要结局是卒中、痴呆及其亚型，以及SVD标志物。根据年龄、性别、种族和其他血管风险因素进行调整。在孟德尔随机化（MR）分析中，报告了每个遗传评分与临床或MRI终点相关的OR或风险比。
Results
Among 502,383 participants (mean [SD] age,56.5[8.1] years;54.4%female), there were 7,668 cases of all-cause dementia and 10,334 strokes. In longitudinal analyses, after controlling for cardiovascular risk factors, participants with insomnia, daytime napping, and dozing were associated with increased risk of any stroke (HR 1.05,95%CI 1.01–1.11, p = 8.53 ×10−3; HR 1.09,95%CI 1.05–1.14, p = 3.20 ×10−;HR 1.19,95%CI 1.08–1.32, p = 4.89 ×10−,respectively). Almost all sleep measures were associated with dementia risk (all p < 0.001, except insomnia). Cross-sectional analyses identified associations between napping, snoring, and MRI markers of SVD (all p < 0.001). MR analyses supported a causal link between genetically predicted insomnia and increased stroke risk (OR 1.31,95%CI 1.13–1.51, p = 0.00072）， but not with dementia or SVD markers.
结果
在502,383名参与者中（平均年龄56.5岁，标准差8.1岁；女性占54.4%），记录了7,668例全因性痴呆和10,334例卒中。在纵向分析中，控制了心血管风险因素之后，失眠、白天小憩以及打瞌睡的参与者与任何卒中的风险增加有关（HR 1.05，95%CI 1.01–1.11，p = 8.53 ×10−3；HR 1.09，95%CI 1.05–1.14，p = 3.20 ×10−3；HR 1.19，95%CI 1.08–1.32，p = 4.89 ×10−4，分别）。几乎所有睡眠相关的测量指标都与痴呆风险增加有关（除了失眠，其余全为p < 0.001）。横断面分析发现，午睡、打鼾与MRI中小血管病（SVD）标志物相关（均为p < 0.001）。孟德尔随机化分析表明基因预测的失眠与卒中风险增加之间存在因果关系（OR 1.31，95%CI 1.13–1.51，p = 0.00072），但与痴呆或SVD标志物之间没有发现因果关系。
Discussion
We found that multiple sleep measures predicted future risk of stroke and dementia, but these associations were attenuated after controlling for cardiovascular risk factors and were absent in MR analyses for Alzheimer disease. This suggests possible confounding or reverse causation, implying caution before proposing sleep disorder modifications for dementia treatment.
讨论
我们发现，多种睡眠指标能够预测未来的卒中和痴呆风险，但在控制心血管风险因素后，这些关联减弱，并且在阿尔茨海默病的孟德尔随机化分析中未能显示关联。这暗示了可能存在混杂因素或反向因果关系，因此在建议通过调整睡眠障碍来治疗痴呆症之前应谨慎。
Introduction
Sleep disorders have been suggested as a causal risk factor of both stroke and dementia.1,2 Studies investigating associa-
tions of sleep have investigated several sleep phenotypes including sleep duration, sleep chronotype, insomnia, napping,daytime dozing, and snoring. Many associations with these measures have been reported: Both short and long sleep duration are associated with increased risk of overall cardio-
vascular disease mortality,3 insomnia is associated with cerebral small vessel disease (SVD) risk,4 and both short sleep duration and insomnia are associated with increased dementia risk.5,6
介绍
睡眠障碍已被认为是卒中和痴呆的潜在风险因素。研究已经调查了多种睡眠特征，包括睡眠时间、睡眠节律、失眠、午睡、白天打瞌睡和打鼾。许多研究报告显示：无论是短时间还是长时间的睡眠，都与总体心血管疾病死亡率的风险增加相关；失眠与脑小血管病（SVD）风险相关；而短时间的睡眠和失眠都与痴呆风险增加相关。
Most previous studies investigating sleep associations with stroke and dementia have been cross-sectional, which leaves
open questions regarding causality. The previously identified associations with sleep could be due to confounding with vascular risk factors such as smoking and alcohol, or could arise from reverse causation, in which subclinical cerebro vascular disease or dementia causes sleep disturbance.
Evidence supporting a causal relationship can be obtained from longitudinal studies evaluating whether risk factors at
baseline predict incident stroke and dementia and by using Mendelian randomization (MR). MR is a statistical approach that uses genetic variants as instrumental variables to infer the
causal effect of a modifiable exposure (risk factor) on a health outcome (e.g., stroke or dementia).
大多数之前研究睡眠与卒中及痴呆关联的研究都是横断面的，这留下了因果关系的问题。之前识别出的睡眠与卒中或痴呆的关联可能由于与吸烟、饮酒等心血管风险因素的混杂，或者可能源于反向因果关系，即亚临床脑血管疾病或痴呆导致睡眠障碍。
通过评估基线风险因素是否能预测未来发生的卒中和痴呆的纵向研究，以及使用孟德尔随机化（MR）分析，可以获得支持因果关系的证据。MR是一种统计方法，利用基因变异作为工具变量来推断可调节暴露（风险因素）对健康结果（例如卒中或痴呆）的因果效应。
To investigate the role of sleep characteristics on stroke and dementia risk, we performed longitudinal analyses in over
500,000 individuals from UK Biobank and examined whether 6 different sleep measures predicted incident stroke (all stroke [AS], ischemic stroke, intracerebral hemorrhage
[ICH]) and dementia (all-cause dementia, Alzheimer disease [AD], vascular dementia, frontotemporal dementia [FTD]). We then performed MR to analyze the causal nature of theassociations.
A recent hypothesis is that disruption of the glymphatic system may play a key role in SVD, which is a major cause of lacunar stroke, ICH, and vascular dementia. This suggests that sleep disorders might specifically increase SVD risk. To investigate, we also evaluated in 40,000 UK
Biobank participants with brain imaging whether sleep measures were
associated with MRI features of SVD, including white matter hyperintensities (WMHs) and markers of white matter ultrastructural damage on diffusion tensor imaging (DTI). We further examined associations between sleep and MRI
markers of SVD using MR.
为了研究睡眠特征对卒中和痴呆风险的影响，我们在超过500000名参与者中进行了纵向分析，检查了6种不同的睡眠指标是否能预测发生的卒中（所有卒中[AS]、缺血性卒中、脑出血[ICH]）和痴呆（全因性痴呆、阿尔茨海默病[AD]、血管性痴呆、额颞叶痴呆[FTD]）。然后，我们进行了孟德尔随机化分析，以分析这些关联的因果性质。
有一种假设认为，糖淋巴系统的破坏可能在脑小血管病（SVD）中发挥重要作用，而SVD是腔隙性卒中、脑出血和血管性痴呆的主要原因。这提示睡眠障碍可能特异性地增加SVD的风险。为此，我们在40000名参与者中评估了睡眠指标与MRI中SVD特征的关联，包括白质高信号（WMHs）和扩散张量成像（DTI）中的白质超微结构损伤标志物。我们进一步使用孟德尔随机化分析了睡眠与MRI标志物的关联。
Methods
Study Population
The UK Biobank is a prospective cohort study of 502,383 participants (aged 40–69 years) recruited from 22 centers across the United Kingdom from March 2006 to October
2010.8 Participants completed self-reported questionnaires, verbal interviews, physical measurements, and blood sample
collection. All UK Biobank protocols were approved by external ethics committees (reference 11/NW/0382), and all participants provided informed consent.9
Sleep Measures
The UK Biobank recorded information using the touchscreen questionnaire on several sleep measures including sleep duration, chronotype (morning/evening person), daytime napping (short periods of sleep taken throughout the day), sleeplessness/insomnia (trouble falling asleep at night or wake up in the middle of night), daytime dozing (inability to stay awake and alert during waking hours), and snoring.
方法
研究人群
本研究是一个前瞻性队列研究，共招募了502,383名年龄在40至69岁之间的参与者，来自英国22个中心，招募时间为2006年3月至2010年10月。参与者完成了自我报告问卷、口头访谈、身体测量和血样采集。所有UK Biobank的研究协议均获得了外部伦理委员会的批准（参考编号11/NW/0382），所有参与者均提供了知情同意。
睡眠指标
本研究通过触摸屏问卷记录了多种睡眠指标的信息，包括睡眠时长、睡眠节律（早型/晚型）、白天小睡（全天的小段睡眠）、失眠（夜间入睡困难或半夜觉醒）、白天打瞌睡（白天难以保持清醒和警觉）以及打鼾。
Incident Stroke and Dementia
Clinical end points were recorded for AS, ischemic stroke, ICH, all-cause dementia, AD, vascular dementia, and FTD. These were defined based on the earliest recorded case that occurred after baseline assessment and before the end of the follow-up period. Data were obtained through self-report at a nurse interview and linkage to hospital admissions from electronic health records and death certificate records. The lists of clinical codes used to define the clinical end points were developed and validated by the UK Biobank Outcome Adjudication Group in conjunction with clinical experts.
MRI Markers of SVD
In 2014, the UK Biobank commenced an imaging study to conduct MRI scans in a subset of;100,000 participants.10 Information for over 40,000 participants had been released at the time of this analysis. Patients with prevalent stroke were excluded. We examined WMH volume and several DTI metrics, including mean diffusivity (MD) and fractional anisotropy (FA).11 We log-transformed the total volume of WMHs from T1 and T2 fluid-attenuated inversion recovery images. For the DTI metrics, we performed principal component analyses on 48 markers ofFA and MD derived by the UK Biobank from the FA skeleton of the diffusion MRI data, and we used the first principal component from each analysis as a summary measure.12 In addition, from DTI scans, we calculated peak width of skeletonized mean diffusivity (PSMD), an automated measure based on skeletonization analysis, using a published pipeline.13
事件性卒中和痴呆
临床终点记录了所有卒中（AS）、缺血性卒中、脑内出血（ICH）、全因性痴呆、阿尔茨海默病（AD）、血管性痴呆和额颞叶痴呆（FTD）。这些终点的定义是基于基线评估后以及随访期结束前记录的最早病例。数据通过护士访谈中的自我报告和电子健康记录以及死亡证明记录中的医院入院数据进行获取。用于定义临床终点的临床代码列表由UK Biobank结果裁定组与临床专家共同制定并验证。
SVD的MRI标志物
2014年，UK Biobank启动了一项成像研究，对大约100,000名参与者进行了MRI扫描。在本次分析时，已有超过40,000名参与者的信息被公开。已有卒中的患者被排除。我们检查了白质高信号（WMH）的体积和几个扩散张量成像（DTI）指标，包括平均扩散率（MD）和分数各向异性（FA）。我们对T1和T2流体衰减反转恢复图像中的WMH总量进行了对数变换。对于DTI指标，我们对UK Biobank从扩散MRI数据的FA骨架中得出的48个FA和MD标志物进行了主成分分析，并使用每个分析的第一个主成分作为汇总指标。此外，我们还计算了从DTI扫描中得出的骨架化平均扩散率的峰值宽度（PSMD），这是一种基于骨架化分析的自动化测量方法，使用了已发布的处理流程。
Genetic Instruments
Sleep measures were used as instrumental variables. We obtained genome-wide association study (GWAS) summary
statistics from published analyses of UK Biobank participants for sleep duration (N = 446,118;78loci),14 chronotype (N = 697,828;351loci),15 daytime napping (N = 452,633;123loci),15 daytime dozing (N = 452,071;42loci),16 and snoring (N = 314,449;41loci).17 For insomnia, the most recent and largest GWAS summary statistics were used, with 554 genetic loci identified in 2,365,010 individuals.18
遗传工具
睡眠指标被用作工具变量。我们从已发布的UK Biobank参与者分析中获得了全基因组关联研究（GWAS）总结统计数据，涵盖了睡眠时长（N = 446,118；78个位点）、睡眠节律（N = 697,828；351个位点）、白天午睡（N = 452,633；123个位点）、白天打瞌睡（N = 452,071；42个位点）和打鼾（N = 314,449；41个位点）。对于失眠，我们使用了最新且最大规模的GWAS总结统计数据，涉及2,365,010名个体，识别出554个遗传位点。
For outcome variables, summary statistics for stroke and is chemic stroke subtypes were obtained from participants of European ancestry from the GIGASTROKE Consortium,19 which consisted of 73,652 patients with stroke and 1,234,808
controls. We conducted analyses for any stroke (n = 73,652 cases), ischemic stroke (n = 62,100 cases), cardioembolic stroke (n = 10,804 cases), large-artery stroke (n = 6,399), and small vessel stroke (SVS; n = 6,811）。 We also used summary statistics from a cohort of neuroimaging-confirmed lacunar (small vessel) stroke, which provided more detailed phenotyping of SVS (NC_SVS, N = 6,030）。20 Summary statistics for AD were obtained from the International Genomics of Alzheimer’s Project (IGAP) (N = 21,982).21 For SVD imaging traits, summary statistics for WMH (N = 42,310), FA (N = 17,663), and MD (N = 17,467) were obtained from a GWAS of participants from the UK Biobank and the CHARGE Consortium.22 We obtained summary statistics for PSMD
from a currently unpublished GWAS (N = 40,464). A summary for originating GWASs is provided in eTable 1
对于结果变量，我们从GIGASTROKE联盟中获得了卒中及其亚型的总结统计数据，该联盟由73,652名卒中患者和1,234,808名对照组成。我们分析了所有卒中（n = 73,652例）、缺血性卒中（n = 62,100例）、心源性卒中（n = 10,804例）、大动脉卒中（n = 6,399例）和小血管卒中（SVS；n = 6,811例）。我们还使用了神经影像学确认的腔隙性（小血管）卒中的总结统计数据，这些数据提供了对SVS更详细的表型信息（NC_SVS，N = 6,030）。关于AD的总结统计数据来自国际阿尔茨海默病项目（IGAP）（N = 21,982）。对于SVD影像学特征，我们从UK Biobank和CHARGE联盟的GWAS中获得了WMH（N = 42,310）、FA（N = 17,663）和MD（N = 17,467）的总结统计数据。我们从当前未公开的GWAS中获得了PSMD的总结统计数据（N = 40,464）。GWAS的总结信息详见eTable 1。
Statistical Analyses
Cross-Sectional and Longitudinal Analyses Ethnicity, smoking, and alcohol were coded as binary outcomes. For ethnicity, European was encoded to “0” and all other ethnicities were encoded to “1.” For smoking and alcohol, “never” was encoded to “0” while “previous” and “current” were encoded to “1.” Categorical sleep variables
(insomnia, chronotype, napping, dozing) were also reconstructed as binary outcomes. For insomnia,“never/rarely” was encoded to “0” while “sometimes” and “usually” were enco
ded to “1.” For daytime napping and dozing, “never/rarely” was encoded to “0” while “sometimes” and “often” were encoded to “1.” For chronotype, “definitely a ‘morning’ per
son” and “more a ‘morning’ than ‘evening’ person” were encoded to 0, whereas “definitely an ‘evening’ person” and
“more an ‘evening’ than ‘morning’ person” were encoded to 1. The associations between sleep measures and SVD imaging
markers (WMH, FA, MD, PSMD) were examined in a crosssectional analysis using linear regression models. WMH was
log-transformed and the continuous outcomes were rescaled to have a mean of 0 and SD of 1. In primary analyses, we only
adjusted for age, sex, and ethnicity. In secondary analyses, we adjusted for a wider range of vascular risk factors and other
potential confounders (age, sex, ethnicity, body mass index, blood pressure treatment, systolic blood pressure, diastolic blood pressure, type 2 diabetes, smoking, alcohol, serum cholesterol, and Townsend deprivation index). Sensitivity analyses were performed with further adjustment for major depression and atypical antipsychotic medication usage.
横断面和纵向分析
种族、吸烟和饮酒被编码为二元变量。对于种族，欧洲裔被编码为“0”，其他所有种族被编码为“1”。对于吸烟和饮酒，"从未"被编码为“0”，而"以前"和"现在"则被编码为“1”。分类的睡眠变量（失眠、睡眠节律、午睡、打瞌睡）也被重新编码为二元变量。对于失眠，“从未/很少”被编码为“0”，而“有时”和“通常”则被编码为“1”。对于白天午睡和打瞌睡，“从未/很少”被编码为“0”，“有时”和“经常”则被编码为“1”。对于睡眠节律，“绝对是‘早晨型’”和“更偏向‘早晨型’”被编码为“0”，而“绝对是‘夜晚型’”和“更偏向‘夜晚型’”则被编码为“1”。在横断面分析中，我们使用线性回归模型检查了睡眠指标与SVD影像学标志物（WMH、FA、MD、PSMD）之间的关联。WMH数据进行了对数转换，连续变量被重新调整为均值为0，标准差为1。在主要分析中，我们仅调整了年龄、性别和种族。在次要分析中，我们调整了更广泛的血管风险因素和其他潜在混杂因素（年龄、性别、种族、体重指数、血压治疗、收缩压、舒张压、2型糖尿病、吸烟、饮酒、血清胆固醇以及Townsend贫困指数）。敏感性分析则进一步调整了重度抑郁症和非典型抗精神病药物使用情况。
Longitudinal analyses were performed to investigate whether sleep measures predicted incident stroke and dementia. In the
analyses, all prevalent cases (outcomes that occurred before the baseline assessments) were excluded. Cox proportional hazards regression models were used to examine the association between sleep variables and risk of incident stroke and dementia. Primary analyses were conducted with adjustment
for age, sex, and ethnicity.Secon-daryanalyses were conducted
with a wide range of vascular risk factors and potential confounders included as described for the cross-sectional analyses. Additionally, sensitivity analyses were performed with
further adjustment for major depression and atypical antipsychotic medication usage and excluding all outcomes of interest that occurred within 1 year of the baseline assessment. The proportional hazards
assumption was evaluated using Schoenfeld residuals.
我们进行了纵向分析以探讨睡眠指标是否能预测新发卒中和痴呆。在这些分析中，所有基线评估之前发生的现有病例（结果）都被排除。我们采用了Cox比例风险回归模型来研究睡眠变量与新发卒中和痴呆的风险之间的关系。主要分析中调整了年龄、性别和种族。次要分析则包括了广泛的血管风险因素和潜在混杂因素，这些因素在横断面分析中已经说明。此外，敏感性分析中进一步调整了重度抑郁症和非典型抗精神病药物的使用情况，并排除了基线评估后1年内发生的所有感兴趣结果。我们使用Schoenfeld残差来评估比例风险的假设。
MR Analyses
A 2-sample MR analysis was performed to examine whether there was evidence to support a causal relationship of sleep
with stroke, dementia, and SVD imaging markers. Primary analyses were conducted using 2-sample inverse variance weighted univariable MR (IVW-MR). Independent
genetic variants (r 2 < 0.01 in European ancestry individuals in the 1000 Genomes Project, Phase 3 release [1KG]) that were
associated with sleep measures at genome-wide significance (p < 5 ×10−8） were selected in European ancestry individuals.
These variants were cross-referenced against the PhenoScanner database of published genetic associations to ensure
that they, or their proxies (r 2 ≥ 0.8 in the 1KG project), were not associated with potential confounding factors at genome wide significance.23 Details of the excluded genetic variants are listed in eTable 2 (links.lww.com/WNL/D422). For all
analyses, palindromic variants with ambiguous allele frequencies were discarded as were genetic variants with potential strand issues that could not be resolved. Furthermore, all variants associated with sleep measures were harmonized with the outcome data to ensure that the effect estimates of each variant on sleep and the outcome corresponded to the same-effect allele.
MR分析
进行了双样本孟德尔随机化（MR）分析，以探讨睡眠是否与卒中、痴呆和SVD影像标志物之间存在因果关系的证据。主要分析使用了双样本逆方差加权单变量MR（IVW-MR）。我们选择了在全基因组显著性水平（p < 5 ×10^-8）下与睡眠指标相关的独立遗传变异（在1000基因组计划第3阶段发布中的欧洲祖先个体中r^2 < 0.01）。这些变异与PhenoScanner数据库中已发布的遗传关联进行了交叉参考，以确保它们或其代理（在1KG项目中r^2 ≥ 0.8）不与潜在的混杂因素在全基因组显著性水平相关。23排除的遗传变异的详细信息列在eTable 2中。在所有分析中，含有模棱两可等位基因频率的回文变异以及可能存在无法解决的链问题的遗传变异均被舍弃。此外，与睡眠指标相关的所有变异都与结果数据进行了协调，以确保每个变异对睡眠和结果的效应估计对应相同的效应等位基因。
A range of sensitivity analyses were performed relaxing some of the stricter assumptions underlying the IVW-MR method,
including the weighted median estimator, simple and weighted mode-based estimators, MR-Egger regression, and MR pleiotropy residual sum and outlier (MR-PRESSO)
methods.24 These methods are recommended in practice for sensitivity analyses because they require different assumptions to be satisfied, and therefore, if estimates from such methods are similar, any inferred causal claims are more credible.25 MR pleiotropy tests were performed which examine the intercept term in MR-Egger regression to evaluate whether the result is
influenced by directional horizontal pleiotropy. For significant results, reverse MR was performed to examine whether it is the case that the outcome causes the exposure. MR-PRESSO distortion and outlier tests were performed.
All observational analyses were conducted using Python 3.92 software; MR analyses were conducted in R v4.2.0 using
the TwoSampleMR and MR-PRESSO packages. A false discovery rate (FDR) correction was applied with a significance
threshold of 0.05.
进行了多种敏感性分析，放宽了IVW-MR方法中一些严格的假设，包括加权中位数估计器、简单和加权模式估计器、MR-Egger回归以及MR多重性残差和离群值（MR-PRESSO）方法。24 这些方法在实际操作中被推荐用于敏感性分析，因为它们要求满足不同的假设，因此如果这些方法的估计结果相似，任何推断的因果关系就更具可信度。25 进行了MR多重性测试，以检查MR-Egger回归中的截距项，以评估结果是否受到定向水平多重性的影响。对于显著结果，进行了逆向MR分析，以检查结果是否是由暴露因素引起的。还进行了MR-PRESSO的失真和离群值测试。
所有观察性分析使用Python 3.92软件进行；MR分析则使用R v4.2.0软件中的TwoSampleMR和MR-PRESSO包进行。采用了0.05的显著性阈值进行了虚假发现率（FDR）校正。
Standard Protocol Approvals, Registrations, and Patient Consents
All UK Biobank participants provided informed consent as part of the UK Biobank recruitment process to the use of their
anonymized data and samples for any health-related research, to be recontacted for further substudies, and for UK Biobank
to access their electronic health records. The UK Biobank has approval from the North West Multi-centre Research Ethics
Committee as a Research Tissue Bank approval. This research was conducted using UK Biobank under application number
36509.
Data Availability
The data supporting the findings of this study are available within the article and its Supplemental Materials. The original
data from UK Biobank can be accessed by approved researchers through application to UK Biobank (ukbiobank.ac. uk/enable-your-research). The summary statistics obtained
from the genome-wide association are publicly available. The summary statistics for sleep characteristics can be obtained
from the Sleep Disorder Knowledge Portal (sleep.hugeamp. org/) and the Complex Traits Genetics Lab (ctg.cncr.nl/
software/summary_statistics/). The summary statistics for stroke from the GIGASTROKE Consortium can be obtained from the GWAS Catalog (ebi.ac.uk/gwas/; study accession
numbers GCST90104534–GCST90104563). The summary statistics for AD can be
obtained from the IGAP (niagads.org/
datasets/ng00075).
标准协议批准、注册和患者同意
所有UK Biobank参与者在UK Biobank招募过程中均已提供知情同意，允许使用其匿名数据和样本进行任何健康相关研究，接受进一步的子研究联系，并允许UK Biobank访问其电子健康记录。UK Biobank已获得北西部多中心研究伦理委员会的研究组织样本库批准。本研究使用了UK Biobank，申请编号为36509。
数据可用性
本研究的发现数据可以在文章及其补充材料中找到。UK Biobank的原始数据可以通过向UK Biobank申请（ukbiobank.ac.uk/enable-your-research）由获批准的研究人员访问。从全基因组关联研究获得的汇总统计数据是公开可用的。睡眠特征的汇总统计数据可以从睡眠障碍知识门户网站（sleep.hugeamp.org/）和复杂性状遗传学实验室（ctg.cncr.nl/software/summary_statistics/）获取。来自GIGASTROKE联盟的卒中汇总统计数据可以从GWAS目录（ebi.ac.uk/gwas/; 研究申请编号GCST90104534–GCST90104563）获取。AD的汇总统计数据可以从IGAP（niagads.org/datasets/ng00075）获取。
Results
Participant Characteristics
A total of 502,383 participants from the UK Biobank were analyzed, with mean age 56.5 (SD 8.1) years; 54.4% were female and 94.6% White. Details for other risk factors, sleep measures, and SVD imaging measures are summarized in Table 1.
Association of Sleep Measures With Stroke and Dementia: Longitudinal Analysis
The median number of years of follow-up was 13.07 (interquartile range [IQR] 1.40) for stroke and 13.08 (IQR 1.39) for dementia. During the follow-up period, there were 10,434 （2.07%） incident strokes（1.7%） ischemic strokes,1,859（0.37%） cases of ICH（1.5%） cases of all-cause dementia（0.6%） cases of AD（0.34%） cases of
vascular dementia, and 264（0.05%） cases of FTD (Table 2).
结果
参与者特征
本研究分析了502,383名来自UK Biobank的参与者，平均年龄为56.5岁（标准差8.1岁）；其中54.4%为女性，94.6%为白人。其他风险因素、睡眠指标和SVD影像指标的详细信息总结见表1。
睡眠指标与卒中和痴呆的关联：纵向分析
卒中的随访中位年限为13.07年（四分位数范围1.40年），痴呆的随访中位年限为13.08年（四分位数范围1.39年）。在随访期间，共发生了10,434例（2.07%）新发卒中，其中1.7%为缺血性卒中；1,859例（0.37%）为脑出血（ICH）；所有原因的痴呆发生了1,359例（0.6%）；阿尔茨海默病（AD）发生了1,075例（0.34%）；血管性痴呆发生了264例（0.05%）；前额颞叶痴呆（FTD）发生了264例（0.05%）（表2）。
All sleep measures, except insomnia, were associated with all cause dementia, and all sleep measures, except snoring, were
associated with vascular dementia. After adjusting for vascular risk factors, all associations with all-cause dementia remained statistically significant (all p < 0.01), apart from insomnia,which was no longer significant. The association of daytime
dozing with vascular dementia attenuated but remained statistically significant after adjusting for vascular risk factors (hazard ratio [HR] 1.36 [1.10–1.69], p = 0.005). Finally, we found that insomnia (HR0.83 [0.77–0.90], p = 9.50 × 10−6) and snoring (HR 1.11 [1.02–1.21], p = 0.008) were signifi-
cantly associated with AD both before and after adjusting for cardiovascular risk factors.
In longitudinal analyses accounting for age, sex, and ethnicity, there were strong associations between insomnia, chro-
notype, daytime napping, and dozing with AS and all ischemic stroke (AIS). However, after adjusting for vascular risk factors,
only insomnia, daytime napping, and dozing showed weak associations with AS (HR 1.05 [1.01–1.11], p = 8.53 × 10−3 for insomnia; HR 1.09 [1.05–1.14], p = 3.20 × 10−5 for
napping; HR 1.19 [1.08–1.32], p = 4.89 × 10−4 for dozing) and AIS (HR 1.08 [1.03–1.13], p = 0.002 for insomnia; HR
1.11 [1.06–1.16], p = 4.43 × 10−6 for napping; HR 1.24 [1.11–1.37], p = 6.64 × 10−5 for dozing). The sensitivity analyses with further adjustment for major depression and atypical antipsychotics medication usage and excluding all outcomes of interest that occurred within 1 year of the baseline assessment did not affect any significance level. To evaluate validity ofmodels, scaled Schoenfeld residuals global tests were performed.
所有睡眠指标中，除了失眠以外，都与全因痴呆相关；所有睡眠指标中，除了打鼾以外，都与血管性痴呆相关。调整血管风险因素后，除了失眠外，所有与全因痴呆的关联依然具有统计学意义（均 p < 0.01），而失眠的关联则不再显著。调整血管风险因素后，白天打瞌睡与血管性痴呆的关联有所减弱，但仍保持统计学显著性（风险比 [HR] 1.36 [1.10–1.69], p = 0.005）。最后，我们发现失眠（HR 0.83 [0.77–0.90], p = 9.50 × 10−6）和打鼾（HR 1.11 [1.02–1.21], p = 0.008）在调整心血管风险因素之前和之后均显著与阿尔茨海默病（AD）相关。
在考虑年龄、性别和种族的纵向分析中，失眠、睡眠时间类型、白天打盹和打瞌睡与所有卒中（AS）和所有缺血性卒中（AIS）之间存在强关联。然而，调整血管风险因素后，只有失眠、白天打盹和打瞌睡与AS（失眠 HR 1.05 [1.01–1.11], p = 8.53 × 10−3；打盹 HR 1.09 [1.05–1.14], p = 3.20 × 10−5；打瞌睡 HR 1.19 [1.08–1.32], p = 4.89 × 10−4）以及AIS（失眠 HR 1.08 [1.03–1.13], p = 0.002；打盹 HR 1.11 [1.06–1.16], p = 4.43 × 10−6；打瞌睡 HR 1.24 [1.11–1.37], p = 6.64 × 10−5）显示出较弱的关联。进一步调整主要抑郁症和非典型抗精神病药物使用以及排除基线评估后1年内发生的所有相关结果的敏感性分析未改变任何显著性水平。为了评估模型的有效性，进行了缩放Schoenfeld残差的全局测试。
The results for longitudinal studies after adjusting for confounders and vascular risk factors are presented in Figure 1.Detailed association results are presented in eTable 3 (links.lww.com/WNL/D422) for primary analyses, eTable 4 forsecondary analyses, eTables 5 and 6 for sensitivity analyses,
and eTable 7 for scaled Schoenfeld residual tests.
调整混杂因素和血管风险因素后的纵向研究结果如图1所示。主要分析的详细结果可以在eTable 3（链接: links.lww.com/WNL/D422）中找到，次要分析的结果见eTable 4，敏感性分析的结果在eTables 5和6中，缩放Schoenfeld残差测试的结果在eTable 7中列出。
Association of Sleep Measures With MRI
Markers of SVD: Cross-Sectional Analysis
After adjustment for sex, age, and ethnicity, daytime napping was associated with higher WMH (odds ratio [OR] 1.99, p =
5.56 × 10−13 [1.95–2.03]), FA(OR1.24 [1.14–1.35], p = 7.21×10−7), MD (OR 1.17 [1.07.1.27], p = 2.82 × 10−4), and PSMD (OR 1.05 [1.03–1.07], p = 2.26 × 10−8). Furthermore,insomnia (OR 1.16 [1.06–1.27], p = 0.002) and snoring (OR 0.97 [0.78–0.93], p = 0.004) were associated with FA while
chronotype (OR 0.89 [0.81–0.97], p = 0.006) and snoring (OR 1.12 [1.03–1.22], p = 0.01) were associated with MD (Table 3 and eTable 8, links.lww.com/WNL/D422). After further adjusting for vascular risk factors, napping showed strong evidence of a weak association with WMH (OR 1.05 [1.03–1.07], p = 1.49 × 10−6), FA (OR 1.13 [1.04–1.23], p =0.005), MD (OR 1.14 [1.05–1.24], p = 0.002), and PSMD (OR 1.03 [1.02–1.05], p = 2.84 × 10−4) (eTable 9). People with an evening chronotype were more likely to have a higher MD (OR 0.9 [0.83–0.98], p = 0.017). However, other associations were no longer statistically significant. Sensitivity analyses with further adjustment for major depression and atypical antipsychotics medication usage showed similar results (eTable 10).
睡眠指标与SVD的MRI标记的关联：横断面分析
在调整了性别、年龄和种族因素后，白质高信号（WMH）与白质纤维的各向异性（FA）、平均扩散率（MD）以及骨架化平均扩散率（PSMD）之间存在显著关联。具体而言，白天打瞌睡与较高的WMH（比值比 [OR] 1.99, p = 5.56 × 10^-13 [1.95–2.03]）、FA（OR 1.24 [1.14–1.35], p = 7.21 × 10^-7）、MD（OR 1.17 [1.07–1.27], p = 2.82 × 10^-4）和PSMD（OR 1.05 [1.03–1.07], p = 2.26 × 10^-8）相关。此外，失眠（OR 1.16 [1.06–1.27], p = 0.002）和打鼾（OR 0.97 [0.78–0.93], p = 0.004）与FA相关，而生物钟类型（OR 0.89 [0.81–0.97], p = 0.006）和打鼾（OR 1.12 [1.03–1.22], p = 0.01）与MD相关（见表3和eTable 8，链接: links.lww.com/WNL/D422）。在进一步调整血管风险因素后，打瞌睡与WMH（OR 1.05 [1.03–1.07], p = 1.49 × 10^-6）、FA（OR 1.13 [1.04–1.23], p = 0.005）、MD（OR 1.14 [1.05–1.24], p = 0.002）和PSMD（OR 1.03 [1.02–1.05], p = 2.84 × 10^-4）之间的弱关联仍然存在（见eTable 9）。夜型生物钟的个体更可能拥有较高的MD（OR 0.9 [0.83–0.98], p = 0.017）。然而，其他关联不再具有统计学意义。进一步调整主要抑郁症和非典型抗精神病药物使用的敏感性分析显示了类似的结果（见eTable 10）。
MR Analyses
MR analyses found no significant association of genetically determined napping and dozing with AS, ischemic stroke, SVS,
neuroimaging-confirmed lacunar stroke, cardioembolic stroke, and AD (all p > 0.2, Figure 2, eTable 11, links.lww.com/WNL/
D422). Genetically elevated propensities for napping and dozing were associated with higher risk of large-artery stroke (OR 1.90
[1.04–3.47], p = 0.035 for napping; OR 3.47 [1.09–16.57], p = 0.037 for dozing), but the results were no longer significant after
FDR correction. Genetically elevated levels of insomnia were associated with increased risk of AS (OR 1.27 [1.10–1.47], p = 0.00072) and AIS (OR 1.31 [1.13–1.51], p = 0.0003) after FDR correction. Insomnia was also significantly associated with SVS (OR 1.56 [1.03–2.36], p = 0.03), but not after FDR correction. No reverse causality was observed for either AS (p = 0.56) or AIS
(p = 0.19) (eTable 12).
There were no statistically significant associations between any sleep measures and WMH, FA, or MD. All other results
were not statistically significant. There was no evidence for pleiotropy (eTable 11, links.lww.com/WNL/D422), and the MR-PRESSO distortion test detected 1 outlier for insomnia and AS and 2 outliers between insomnia and ischemic stroke. The outlier-corrected results were still statistically signifi-
cant (OR 1.28 [1.17–1.49], p = 0.0006 for AS; OR 1.32 [1.14–1.53], p = 0.0002 for AIS). All MR-PRESSO global test, outlier test, distortion test, and outlier-corrected results are summarized in eTables 13–16.
孟德尔随机化分析
孟德尔随机化分析未发现遗传性打瞌睡和打盹与动脉硬化性脑卒中（AS）、缺血性脑卒中、小血管性脑卒中（SVS）、神经影像学确认的腔隙性脑卒中、心源性脑卒中以及阿尔茨海默病（AD）之间有显著关联（所有 p > 0.2，见图2、eTable 11）。遗传性打瞌睡和打盹的倾向与大血管脑卒中的风险增加有关（打瞌睡 OR 1.90 [1.04–3.47], p = 0.035；打盹 OR 3.47 [1.09–16.57], p = 0.037），但在FDR校正后结果不再显著。遗传性失眠水平的升高与AS（OR 1.27 [1.10–1.47], p = 0.00072）和AIS（OR 1.31 [1.13–1.51], p = 0.0003）的风险增加有关，FDR校正后依然显著。失眠与小血管性脑卒中的关联在FDR校正后不再显著（OR 1.56 [1.03–2.36], p = 0.03）。AS（p = 0.56）和AIS（p = 0.19）没有发现逆因果关系（见eTable 12）。
在睡眠指标与WMH、FA或MD之间没有统计学显著关联。其他结果也没有显示统计学上的显著性。没有发现多重效应的证据（见eTable 11，且MR-PRESSO失真测试发现了失眠与AS之间的1个异常值以及失眠与缺血性脑卒中之间的2个异常值。异常值校正后的结果仍具有统计学意义（AS OR 1.28 [1.17–1.49], p = 0.0006；AIS OR 1.32 [1.14–1.53], p = 0.0002）。所有MR-PRESSO全局测试、异常值测试、失真测试及异常值校正结果总结见eTables 13–16。
Discussion
We investigated the relationship of sleep with stroke and dementia, using both observational and genetic data, in over
500,000 individuals. Our observational study found associations between multiple sleep measures and both stroke and dementia, as well as associations between napping and snoring with SVD imaging traits. The association of insomnia with stroke was confirmed in our MR analyses.
讨论
我们通过对超过50万人进行观察性和遗传数据分析，探讨了睡眠与卒中和痴呆症之间的关系。观察性研究发现，多个睡眠指标与卒中和痴呆症存在关联，同时，午睡和打鼾与小血管病（SVD）影像特征有关。孟德尔随机化分析确认了失眠与卒中之间的关系。
In longitudinal analyses evaluating whether sleep measures led to incident stroke and dementia, the results revealed that insomnia, chronotype, daytime napping, daytime sleepiness, and snoring were associated with AS and ischemic stroke,
suggesting a strong relationship between sleep and stroke, consistent with previous evidence.26-28 After adjusting for
confounders and vascular risk factors, the associations for insomnia, daytime napping, and dozing attenuated but remained statistically significant, in accordance with previous work which found that daytime napping was associated with increased risk of stroke.29 We also found that sleep duration, chronotype, daytime napping, daytime sleepiness, and snoring were associated with all-cause dementia, which is
consistent with many previous studies associating sleep duration, chronotype, daytime napping, and daytime sleepiness
with dementia risk.30-33 However, after adjusting for vascular risk factors, only the associations with daytime sleepiness
remained statistically significant. This suggests that the relationship between sleep measures and dementia risk may be mediated by conventional cardiovascular risk factors such as blood pressure, smoking, and alcohol consumption. Although we found a statistically significant association of increased insomnia with reduced risk of AD in the longitudinal analyses, there was no evidence to support a causal association because the MR analyses showed only a weak association of genetically
determined insomnia with increased risk of AD which was not statistically significant. Although the analyses were adjusted
for a wide range of potential confounders and vascular risk factors, there could still be other confounders that may have led to the observed association in the longitudinal analyses.
在纵向分析中评估睡眠指标是否导致卒中和痴呆症的结果显示，失眠、生物钟类型、白天午睡、白天嗜睡和打鼾与动脉硬化性卒中和缺血性卒中有关，这表明睡眠与卒中之间存在显著关联，这与先前的研究结果一致。经过调整混杂因素和血管风险因素后，失眠、白天午睡和打瞌睡的关联虽然减弱，但仍保持统计学意义，这与之前的研究发现白天午睡与卒中风险增加有关的结果一致。我们还发现，睡眠时间、生物钟类型、白天午睡、白天嗜睡和打鼾与全因痴呆症相关，这与许多研究将这些睡眠因素与痴呆症风险联系起来的结果一致。然而，在调整了血管风险因素后，只有白天嗜睡的关联仍然显著。这表明睡眠指标与痴呆症风险之间的关系可能受到血压、吸烟和饮酒等传统心血管风险因素的影响。尽管在纵向分析中发现失眠与阿尔茨海默病（AD）风险降低之间有统计学显著关联，但并无证据表明这种关系是因果性的，因为孟德尔随机化分析显示遗传性失眠与阿尔茨海默病风险增加之间的关联很弱，且不具有统计学显著性。尽管分析已控制了广泛的潜在混杂因素和血管风险因素，但仍可能存在其他混杂因素导致了纵向分析中观察到的关联。
In view of recent hypotheses that sleep is a risk factor of SVD and that this could partially mediate the associations between
sleep and dementia, possibly through the glymphatic system, we examined associations between sleep and MRI markers of SVD in over 40,000 individuals with available brain MRI scans.As well as evaluating the conventional marker WMH, we also examined associations with DTI measures of white matter ultrastructure; such measures have been shown to be more
strongly associated with cognitive impairment than WMH.13,34 FA measures directionality of diffusion, MD measures the extent of diffusion, and PSMD is an automated metric that measures MD within the white matter tracts.13 Fewer associations than with stroke and dementia persisted after controlling
for cardiovascular risk factors, although daytime napping was consistently associated with WMH and all DTI measures after
adjusting for vascular risk factors, which aligns with previous findings.35 Snoring was associated with both diffusivity measures, MD and PSMD.27,28 This suggests that napping and snoring may be risk factors of SVD, although the analysis was cross-sectional and needs to be replicated in a longitudinal
study to reduce risk of confounding.
鉴于近期的假设认为睡眠可能是小血管病（SVD）的风险因素，并且这种风险可能通过脑脊液系统部分地介导了睡眠与痴呆症之间的关系，我们在4万多名拥有脑MRI扫描的个体中探讨了睡眠与SVD的MRI标志物之间的关联。除了传统的白质高信号（WMH）标志物外，我们还研究了与白质微结构的扩散张量成像（DTI）指标的关系，这些指标已被证明与认知障碍的关联比WMH更强。FA（分数各向异性）衡量扩散的方向性，MD（平均扩散率）衡量扩散的范围，而PSMD（白质扩散率）是一个自动化指标，用于测量白质束中的MD。虽然在控制心血管风险因素后，与卒中和痴呆症的关联有所减少，但白天午睡与WMH和所有DTI指标的关联在调整血管风险因素后仍然显著，这与之前的发现一致。打鼾与两个扩散指标MD和PSMD有关。这表明午睡和打鼾可能是SVD的风险因素，但由于分析是横断面的，仍需在纵向研究中进行验证，以减少混杂因素的影响。
To further investigate the causal nature of these associations, our MR analyses found evidence of causal associations linking
genetically determined insomnia to risk of stroke and ischemic stroke. Our MR results did not support causal relationships of
genetically determined daytime napping, sleepiness, and snoring with stroke, dementia, and imaging markers, indicating that these relationships may be confounded by other variables. Our MR analysis did not support a causal association between sleep
characteristics and dementia, which is consistent with a previous study.36 There was no evidence of a causal relationship of
genetically determined insomnia with SVD imaging markers.Several reasons might explain the lack of significant associations in the MR analyses for daytime napping and dozing. Confounding factors may have played a role in the identified associations in the observational study. Although our study
included demographics and vascular risk factors, it is possible that other implicit confounders were not accounted for.37 In
addition, the possibility of reverse causality, where stroke survivors or patients with SVD experience increased daytime sleepiness and napping, cannot be ruled out.32 It is possible
that patients surviving from stroke have increased levels of daytime sleepiness and napping.30
为了进一步探讨这些关联的因果性质，我们的孟德尔随机化（MR）分析发现了遗传决定的失眠与卒中和缺血性卒中风险之间的因果关联。然而，我们的MR结果未能支持遗传决定的白天午睡、嗜睡和打鼾与卒中、痴呆症及影像学标志物之间的因果关系，这表明这些关系可能受到其他变量的混杂影响。我们的MR分析没有支持睡眠特征与痴呆症之间的因果关系，这与之前的研究结果一致。遗传决定的失眠与SVD影像学标志物之间没有发现因果关系。对于白天午睡和打瞌睡在MR分析中缺乏显著关联的原因可能有多个。混杂因素可能在观察性研究中影响了已识别的关联。尽管我们的研究包括了人口统计学和血管风险因素，但可能还有其他隐含的混杂因素未被考虑。此外，逆因果关系的可能性，即卒中幸存者或SVD患者出现增加的白天嗜睡和午睡，也不能排除。卒中幸存者可能具有较高的白天嗜睡和午睡水平。
Previous studies have implicated sleep apnea as a risk factor of stroke,38 and snoring may be indicative of sleep apnea syndrome. However, we found no associations of snoring with stroke in either the observational or MR analyses. The previously reported association might be confounded by other cardiovascular comorbidities such as hypertension and type 2 diabetes, and we included controlling for these in our primary analysis. Consistent with this, previous MR studies have reported no association between sleep apnea and stroke,39 supporting potential confounding of the previously reported epidemiologic associations.38
以往的研究已经将睡眠呼吸暂停症与卒中风险相关联，而打鼾可能是睡眠呼吸暂停综合症的一个标志。然而，我们在观察性研究和孟德尔随机化（MR）分析中均未发现打鼾与卒中之间的关联。这一之前报告的关联可能被高血压和2型糖尿病等其他心血管共病混杂了，我们在主要分析中已考虑这些因素。与此一致，之前的MR研究也未发现睡眠呼吸暂停与卒中之间的关系，这支持了之前流行病学研究结果可能存在混杂因素的观点。
Our study has several strengths. The use of UK Biobank enabled a very large sample size of well-characterized individuals with long-term follow-up to be analyzed, of whom
over 40,000 had brain MRI scans available. We also combined both observational and MR analyses to characterize the nature
of the associations and assess causality, which increased the reliability of the findings. Our study is more comprehensive
than previous analyses with respect to its large sample size, the use of multiple sleep variables as exposures, and the inclusion
of multiple outcome variables including stroke, dementia, and SVD markers, which may provide mechanistic insights.
我们的研究有几个优势。利用英国生物库的数据，使我们能够分析一个样本量极大的、特征明确的个体群体，并对其中超过4万名接受过脑MRI扫描的个体进行长期跟踪。我们还结合了观察性分析和MR分析来描述这些关联的性质并评估因果关系，从而提高了研究结果的可靠性。与之前的分析相比，我们的研究在样本量大、使用多种睡眠变量作为暴露变量以及包含卒中、痴呆症和SVD标志物等多个结果变量方面更为全面，这可能为机制研究提供了新的见解。
However, the study also has limitations. First, most of the sleep measures were derived from self-reported questionnaires. Recently, derived accelerometry data, including measures of sleep duration, have been released in the UK Biobank, which provide more precise measures of sleep not subject to recall bias. Future work should use these data and compare the results with this study. Several factors may have also reduced
the ability of MR analyses to identify associations. We used the UK Biobank for the observational analyses and large GWAS
data sets including the UK Biobank and the GIGASTROKE Consortium for the genetic analyses, which provided much higher statistical power. However, some of the datasets used for the genetic associations with the sleep measures and outcomes
were derived at least partially from the UK Biobank. This overlap in participants may have contributed to some degree of
overfitting and weak instrument bias. The GIGASTROKE Consortium included 12% of cases from the UK Biobank for AS. However, previous MR studies that excluded overlapping participants obtained similar results to their main analyses,19 demonstrating that due to the large sample sizes of the respective studies, the bias due to sample overlap is expected to be very small. Another limitation is that each sleep measure has specified a time frame pertaining to the last 4 weeks. Therefore,
it is possible that the questionnaire may not be capturing a long term exposure because changes in sleep quality that occurred months or years before the baseline assessment may be relevant to long-term effects on health outcomes. Moreover, because of the lack of sufficiently large datasets for genetic associations with dementia subtypes, we were only able to conduct MR analyses using AD summary statistics from IGAP; future work should conduct MR analyses on other types of dementias, including vascular dementia and FTD, when these data become available. Although sensitivity tests were performed, the MR results may still be affected by horizontal pleiotropy and reverse causality, which is a technical limitation in this method. Finally,
if the relationships are U-shaped, as was suggested for sleep duration and SVD traits,40 these may not be detected by MR.
这项研究也存在一些局限性。首先，大多数睡眠数据来自自我报告的问卷调查。最近，英国生物库发布了加速度计数据，包括睡眠持续时间的测量，这些数据提供了更准确的睡眠评估，避免了回忆偏差。未来的研究应使用这些数据，并将其结果与本研究进行比较。MR分析的能力也可能受到一些因素的限制。我们在观察性分析中使用了英国生物库的数据，在遗传分析中则使用了包括英国生物库和GIGASTROKE联盟在内的大型全基因组关联研究（GWAS）数据集，这些数据集提供了更高的统计效能。然而，一些用于分析睡眠测量和结果的遗传数据集至少部分来源于英国生物库。这种参与者的重叠可能导致了一定程度的过拟合和工具偏差。GIGASTROKE联盟中包含了12%的英国生物库的AS病例。然而，之前排除重叠参与者的MR研究得到了与主要分析相似的结果，这表明由于样本量大，样本重叠带来的偏差预期非常小。另一个限制是每个睡眠测量都规定了过去4周的时间范围。因此，问卷可能无法捕捉到长期暴露情况，因为在基线评估前几个月或几年的睡眠质量变化可能对长期健康结果有影响。此外，由于缺乏足够大的数据集来分析痴呆症亚型，我们只能使用IGAP的AD总结统计数据进行MR分析；未来的研究应对其他类型的痴呆症，如血管性痴呆和前额叶痴呆（FTD），进行MR分析，待这些数据可用时。尽管进行了敏感性测试，MR结果仍可能受到水平多效性和逆因果关系的影响，这是该方法的技术局限。最后，如果关系呈U型，比如睡眠持续时间和SVD特征的建议，这种情况可能不会被MR检测到。
The findings of this study have important implications for clinical research and practice. For dementia, our observational analyses identified multiple associations of almost all sleep measures with dementia even after controlling for cardiovas cular risk factors, although none persisted in the MR analyses.
This may reflect limitations in the genetic instruments we had available, but raises caution as to the causality of the observa
tional associations. This is important because correction of sleep disorders has been suggested as preventative therapy for
dementia, but our findings highlight that randomized controlled trials are required before routine sleep interventions should be recommended as a proven treatment. For stroke, we found multiple observational associations, but many of these were no longer significant after controlling for cardiovascular risk factors. However, the associations with insomnia and napping persisted after adjustment, and our MR analyses found associations of genetically determined insomnia with stroke risk and of napping with risk of large-artery stroke, supporting causal relationships. This raises the possibility that treating insomnia may reduce stroke risk and recurrence, but again, this needs testing in clinical trials. Finally, our study does not provide strong evidence that sleep disturbances are a strong risk factor of SVD or that there is a major mechanism linking sleep with dementia. Although associations with napping and snoring
were identified on observational studies, our MR analyses did not confirm evidence of causal relationships.
本研究的发现对临床研究和实践具有重要的启示。在痴呆症方面，我们的观察性分析在控制心血管风险因素后，发现几乎所有的睡眠测量指标都与痴呆症有多重关联，但在孟德尔随机化（MR）分析中，这些关联并未得到持续支持。这可能反映了我们使用的遗传工具的限制，但也对观察性关联的因果关系提出了警示。这一点尤为重要，因为已有研究建议将纠正睡眠障碍作为预防痴呆症的治疗方法，但我们的发现强调，在常规推荐睡眠干预作为有效治疗之前，需要进行随机对照试验。对于卒中，我们发现了多重观察性关联，但许多关联在控制心血管风险因素后变得不再显著。然而，失眠和午睡的关联在调整后依然存在，而我们的MR分析发现了遗传性失眠与卒中风险，以及午睡与大动脉卒中风险的关联，支持了因果关系。这表明治疗失眠可能有助于降低卒中风险及其复发，但这仍需通过临床试验来验证。最后，我们的研究没有提供强有力的证据证明睡眠障碍是SVD的主要风险因素，也没有揭示睡眠与痴呆症之间的主要机制。尽管在观察性研究中发现了午睡和打鼾的关联，但我们的MR分析未能确认这些因果关系的证据。
结束语：
神经内科3病区全体，预祝大家万事如意、心想事成！