Study design
We conducted a two-sample MR analysis using public GWAS summary data (https://gwas.mrcieu.ac.uk/). In this analysis, serum uric acid was considered the "exposure", while coronary artery disease (CAD), hypertension, myocardial infarction (MI), heart failure (HF), atrial fibrillation (AF), angina, and coronary heart disease (CHD) were the "outcomes." This approach evaluated the causal relationship between serum uric acid levels and CVD. We assessed heterogeneity through Cochran's Q test and verified causality reliability using sensitivity analyses, including pleiotropy and "leave-one-out" analyses. Informed consent and ethical approval were obtained in the original publications and publicly available databases. The MR Analysis rested on three critical assumptions: (1) IVs are strongly associated with serum uric acid levels. (2) There is no association between IVs and confounding factors. (3) IVs influence CVD solely through serum uric acid (Figure 1).
Data sources
Our systematic analysis utilized GWAS summary statistics from a comprehensive cohort to determine a causal link between serum uric acid and CVD. The serum uric acid dataset (GWAS ID: ebi-a-GCST90018977) comprised 343,836 participants with 19,041,286 SNPs. The analysis included seven CVDs: CAD (42,096 cases and 361 controls), hypertension (55,917 cases and 162,837 controls), MI (14,825 cases and 44,000 controls), HF (47,309 cases and 93,0014 controls), AF (220,68 cases and 116,926 controls), angina (30,025cases and 440,906 controls) and CHD (60,801 cases and 123,504 controls). To minimize population heterogeneity bias, only aggregated data from European populations were included. For detailed information on the GWAS dataset, refer to Table 1.
Selection of the Instrumental Variables (IVs)
To elucidate the causal relationship between serum uric acid and seven CVDs, we identified IVs for serum uric acid. In this study, single nucleotide polymorphisms (SNPs) were selected as IVs based on the following criteria: (1) SNPs significantly associated with serum uric acid (P<5×10-8) were extracted (15); (2) To ensure SNP independence, those in linkage disequilibrium (LD) (r2=0.0001, kb=10000) were excluded (16); (3) SNPs closely related to the seven CVD were eliminated (P>5×10-5) (17); (4) SNPS that were missing or palindromic with moderate allele frequency were removed from the outcome data (18); (5) The F statistic was calculated for each SNP to assess weak instrumental variable influence, discarding those with F<10 (19).
where n is the exposure sample size, and R2 is the proportion of variation in the exposure database explained by SNPs. R2 is calculated as follows: R2=2×(1-MAF)×(MAF×b2) where MAF is the minor allele frequency and β is the allele effect value; (6) SNPs associated with potential confounders (BMI, blood lipid level, diabetes, alcohol consumption) were identified and removed using the PhenoScanner V2 database (http://www.phenoscanner.medschl.cam.ac.uk/) (20); (7) An MR-PRESSO outlier test was conducted to eliminate anomalous SNP (21). Subsequently, MR Analysis was performed with the included SNPs.
Mendelian randomization(MR) analysis
Four distinct methods were employed to estimate the causal effect between serum uric acid and CVD: inverse variance weighting (IVW), MR-Egger, weighted median (WM), and weighted model (22). These methods address varying levels of horizontal pleiotropy, thereby thoroughly assessing the causal relationship between serum uric acid and CVD: (1) IVW, assuming no horizontal pleiotropy among SNPs, amalgamates Wald ratio estimates from each SNP to yield a combined causality estimate with optimal statistical power (23); (2) MR-Egger accounts for the presence of horizontal pleiotropy and provides consistent results under the Instrument Strength Independent of Direct Effect (InSIDE) assumption (24); (3) WM aggregates data from multiple genetic variants into a singular causal estimate, offering consistent effect estimation even if up to 50% of the data originates from invalid IVs (25); (4) The weighted model assigns causal estimates to each genetic variant inversely proportional to its variance (26). Thus, IVW analysis serves as the primary MR method, complemented by the other three techniques to enhance result reliability. Additionally, we applied False Discovery Rate (FDR) correction, establishing statistical significance at an adjusted P-value threshold of 0.05.
Sensitivity analysis
To ensure a robust MR Estimate, several sensitivity analyses were conducted:
(1) The Cochran’s Q test, applied to both the IVW and MR-Egger methods, assessed heterogeneity. A P-value greater than 0.05 indicated an absence of heterogeneity. Additionally, potential heterogeneity was visually appraised using funnel plots (27). (2) The MR-Egger-intercept test evaluated horizontal pleiotropy, with P>0.05 suggesting no such pleiotropy (28). (3) A leave-one-out analysis was performed by sequentially excluding each SNP to determine if a single SNP biased the IVW estimate results (29).
Statistical analysis
All analyses were conducted using R software (version 4.3.0), employing the "Two-Sample-MR" (version 0.5.6), "MR-PRESSO" (version 1.2), and "MendelianRandomization" (version 0.4.3) R packages for Mendelian randomization.