Introduction:
Due to contentious associations between sleep and stroke risk we performed a meta-analysis of cohort studies and utilized Mendelian randomization (MR).

Material and methods:
For the meta-analysis we pooled prospective studies and then reviewed the largest genome-wide association studies regarding self-reported or accelerometer-derived sleep duration with stroke [ischemic (IS), cardioembolic (CES), large artery (LAS), small vessel (SVS)]. Inverse variance weighted method (IVW), weighted median (WM)-based method, MR-Egger and MR-Pleiotropy RESidual Sum and Outlier (PRESSO) were performed. To determine the impact of single nucleotide polymorphisms (SNPs) leave-one-out method was applied.

Results:
Pooled prospective studies demonstrated shorter (<7h) [n=25 studies, I2 = 71.4, p <0.001; risk ratio (RR): 1.18, 95%CI: 1.08-1.30, p <0.001] and longer (>8h) [n=16 studies, I2 = 53.6, p <0.001; RR: 1.38, 95%CI: 1.24-1.53, p <0.001] sleep increased stroke risk (compared with 7-8h), but were subject to high levels of heterogeneity. In MR, self-reported sleep duration had no significant effect on IS (IVW: beta = -0.031, p = 0.747), CES (IVW: beta = -0.039, p = 0.849), LAS (IVW: beta = -0.246, p = 0.328) and SVS (IVW: beta = -0.102, p = 0.667) risk. This was also observed for short and long accelerometer-derived sleep (all p >0.126). Estimated associations had no significant heterogeneity and MR-PRESSO revealed no outliers. There was low likelihood of pleiotropy (all estimations p >0.539) and associations were not driven by single SNPs.

Conclusions:
Meta-analysis revealed shorter and longer sleep increased total stroke risk, but with high heterogeneity. MR analysis showed no causal associations between sleep duration and stroke risk.