Research and design scheme
In our research, MR analysis was employed to evaluate the causal relationship between 731 immune cell counts and AF. For MR causal inference to be valid, it must adhere to three fundamental assumptions: firstly, a strong correlation exists between genetic variation and the exposure of interest; secondly, the association between genetic variation and exposure is independent of confounding factors; thirdly, genetic variation impacts the outcome solely through the exposure pathway[16]. In this study, bidirectional Mendelian Randomization was utilized to rigorously investigate the causative link between immune cell counts and AF.
Sources of exposure and outcome data
For the exposure analysis, our study incorporated a comprehensive panel of 731 immune cell counts. These were derived using aggregated statistics of blood cell traits from Genome-Wide Association Studies (GWAS) conducted by the Blood Cell Consortium (BCX). The BCX data integrates findings from the UK Biobank and an extensive international collaborative project, involving 563,085 participants predominantly of European ancestry. This GWAS provided genetic variants associated with circulating leukocyte subtypes, including white blood cells, monocytes, lymphocytes, neutrophils, eosinophils, and basophils. Additionally, it encompassed lymphocyte subsets such as HLA DR + natural killer (NK) cells, CD4 regulatory T cells (TCD4), NKT cells, CD4 + CD8dim T cells, CD8 + T cells, and B cells[17]. The Genome-Wide Association Study (GWAS) aggregated data on AF from a comprehensive genome-wide association meta-analysis. This integrated dataset comprises 1,030,836 individuals of European ancestry, segregated into 60,620 diagnosed AF cases and 970,216 control participants[18].
Choice of instrumental variable (IV)
The IV significance level for each immune cell counts was set at 1 × 10 − 5[16], and to obtain an independent IV, we aggregated according to the 1000 Genome Project's linkage disequilibrium (LD) reference panel (R2 < 0.001 at a distance of 1,0000 kb)[19]. Based on F > 10, Filter the data and delete the weak tool variables; In conducting reverse MR Analysis, we adopted a more rigorous selection criterion. The threshold for statistical significance was established at 5×10^-8, and the linkage disequilibrium threshold (r^2) was set at 0.001. Subsequent to the exclusion of instrumental variables (IVs) with low F-statistics (< 10)[15], a total of 111 AF IVs were retained for subsequent analysis.
Statistical analysis
To ascertain the causal relationship between 731 Immune cell counts and AF, we primarily employed random effects inverse variance weighting (IVW)[20] and weighted median (WM)[21] methodologies. These analyses were conducted using the MR package, version 0.4.3, to estimate the impact of exposure on outcomes under the validity of the MR hypothesis. The Cochran Q test (p < 0.05) was applied to assess the residual heterogeneity in the IVW model. Furthermore, the MR-Egger[22, 23] intercept test (p < 0.05) was utilized to evaluate potential pleiotropy in causal estimations. Radial MR tests were conducted to identify multi-effect outliers. A leave-one-out (LOO) analysis was performed to investigate whether single nucleotide polymorphisms (SNPs) could induce bias in causal estimations. Various graphical representations, including scatter plots, funnel plots, and forest plots, were used in our analysis. The scatter plot indicated the non-influence of outliers on the results. The funnel plot demonstrated the robustness of the correlation and the absence of heterogeneity. The forest plot elucidated the interaction between the exposure and outcome tools.