The data that support the findings of this study are available from the corresponding author upon reasonable request.
Supposing the causal estimate of MR studies is convincing.Three pivotal assumptions must be met:1) The selected genetic IVs must be powerfully associated with exposure.2)The selection of genetic IVs do not affect outcome independently of exposure (i.e.,horizontal pleiotropy is nonexistent).3)The selected genetic IVs are unrelated to the potential confounders. Figure 1 shows an overview of the current study design.Because the study was based on publicly available databases and published research,it did not require ethical approval and consent from participants.Figure 1 shows an overview ofthe current study design. Ethical approval and consent to participants were not necessary as the study was based on openly available databases and published studies.
In terms of the exposure,the genome-wide association study(GWAS) meta-analysis with 460,099 participants of European ancestry was applied to select IVs for sleep duration.A full description of the study design,participants and quality control (QC) methods have been described in detail previously.UK Biobank received ethical approval from the Research Ethics Committee (REC reference for UK Biobank is 11/NW/0382).
About outcome,genetic data on AS were obtained from United Kingdom Biobank participants,including 361,194subjects with European ancestry (14,334 cases and 346,860 controls) in total,covering 13,586,589 single nucleotide polymorphisms (SNPs).we introduce covariate-adjusted LD score regression (cov-LDSC),a method to accurately estimate genetic heritability (h2g) and its enrichment in both homogenous and admixed populations with summary statistics and in-sample LD estimates.
The full data release contains the cohort of successfully genotyped samples (n=488,377).49,979individuals were genotyped using the UK BiLEVE array and 438,398 using the UK Biobank axiom array.Pre-imputation QC,phasing and imputation are described elsewhere.
Selection of instrumental variables
In the present study,SNPs were defined as IVs. All requested SNPs conformed with the following conditions:1.strongly correlated with exposure based on genome-wide significance;2.having no linkage disequilibrium(LD)(pairwise r2=0.001,window size=10,000kb);3.Without palindromic structures.According to the three mentioned assumptions and above conditions,a total of 65 SNPs were identified.To achieve powerful estimates,we used proxy SNPs with strong LD(r2>0.8) to substitute for the selected SNPs on condition that the corresponding SNPs were unavailable in AS GWAS.The first-stage regression, or F, statistic was used to assess the strength of the instruments and was calculated using the following equation:F=(R2/k)/([1−R2]/[n−k−1]),where R2 is the proportion of the sleep duration variability accounted for by the SNP, k is the number of instruments used in the model and n is the sample size.To limit the influence of possible weak instrumental variable bias, an F statistic above 10 was expected to be of sufficient strength for the main study.Figure 2 displayed the flow chart of IVs selection.
The inverse-variance weighted(IVW) method was conducted,as the primary method to evaluate the causal association between sleep duration and AS.We chose a fixed-effects model when the p-value,as the result of Cochran’s Q test,is >0.05,otherwise the random-effects model was applied. The IVW method was perceived as the most dependable if the selected IVs did not have directional pleiotropy (p-value for MR-Egger intercept >0.05).
In sensitivity analyses,we chose MR-Egger method to evaluate the potential pleiotropy effects.The MR-Egger regression estimated the causal effect as the slope from the weighted regression of the IVs-outcome associations on the IVs-exposure association,and the intercept term reflected the average pleiotropic effect[23,24]. Additionally,we also applied simple median,weighted median,RadialMR and MR-PRESSO(Mendelian Randomization Pleiotropy Residual Sum and Outlier) outlier test methods to assess the presence of pleiotropy.If more than 50% SNPs are effective IVs,the consistent estimates of the causal effect would be provided by the weighted median. Not only does MR-PRESSO detect pleiotropy,but also it can exclude the outlying SNPs and reassess the effect estimates.Meanwhile,leave-one-out analysis was performed to test the influence of outlying values.To remove the effect of other confounders,we also explored the pleiotropy of each selected SNPs at the GWAS threshold of statistical significance(p-value<5×10−8) by the PhenoScanner V2 database(http://www.phenoscanner.medschl.cam.ac.uk/).
All tests were two sided,and differences were considered as statistical significance (p-value <0.05),unless noted.All of the analyses were conducted using Two Sample MR (V 0.5.6),RadialMR and MR-PRESSO (V 1.0)packages in R software (4.0.5).