Data for alcohol abuse (PubMed ID: 34140656) and smoking (PubMed ID: 34140656) were obtained from the genome-wide association studies (GWAS) through the Social Science Genetic Association Consortium website ( https://www.thessgac.org ). The largest sample of GWAS data on primary open-angle glaucoma (PubMed ID: 29452408) was accessed via the IEU OpenGWAS Project website ( https://gwas.mrcieu.ac.uk ) on April 28, 2024. The study focuses on the potential impact of lifestyle habits of Chinese POAG patients on the condition of glaucoma. However, due to the lack of high-quality GWAS data on lifestyle habits in the Chinese population, the alcohol and smoking GWAS datasets used were from European populations, while the open-angle glaucoma GWAS data originated from East Asia. The alcohol dataset included 6,483,125 single nucleotide polymorphisms (SNPs), the smoking dataset included 1,157,911 SNPs, and the POAG dataset included 5,961,428 SNPs.

1) The instrumental variable must be highly correlated with the exposure, with an F-statistic > 10 indicating significant correlation [12] .

2) The instrumental variable should not be directly associated with the outcome except through the exposure, indicating no pleiotropy. Pleiotropy tests are carried out where P ≥ 0.05 suggests no genetic pleiotropy.

3) The instrumental variables should be independent of unmeasured confounders. Given that the Mendelian randomization approach selects SNPs that adhere to the genetic principle of random allocation of alleles from parents to offspring, they are minimally susceptible to environmental influences and postnatal factors. Thus, it is theoretically assumed that the instrumental variables are independent of environmental factors such as socioeconomic and cultural influences.

In this Mendelian randomization study, alcohol and smoking were analyzed separately. Initially, SNPs with a significance level of P < 5 × 10⁻⁸ were selected from the summary data of alcohol and smoking GWAS to ensure statistical significance. To avoid confounding due to linkage disequilibrium, a linkage disequilibrium coefficient (r²) threshold of 0.001 was set, with a linkage disequilibrium region width limited to 10,000 kb, ensuring the selected SNPs are independent. Subsequently, loci associated with the selected alcohol and smoking-related SNPs were extracted from the summary data of the POAG GWAS, setting a minimum r² of more than 0.8 to ensure high linkage disequilibrium with disease marker loci. SNPs missing in the data were not replaced but directly excluded. Finally, SNPs directly and significantly associated with POAG (P < 5 × 10⁻⁸) [13] were excluded from the consolidated dataset to minimize interference from direct genetic effects, thus precisely assessing the potential impact of alcohol abuse and smoking on glaucoma risk. The specific selection process is illustrated in Figure 1 .

To validate the causal relationship between the exposure factors and POAG, this study employs three Mendelian randomization regression models: MR-Egger regression, Weighted Median Estimator (WME), and Inverse-Variance Weighted (IVW) method. These methods use SNPs as instrumental variables. 1) MR-Egger regression analyzes the relationship between each SNP with the exposure and outcome, fitting a linear regression model to estimate a non-zero intercept, thereby detecting potential instrumental variable bias. 2) The WME method calculates the causal effect estimates (β_j) for each SNP and provides a robust optimization of the causal relationship using the median of all valid SNPs. 3) The IVW method relies on summary data to calculate the weighted average of the associations between each SNP and the exposure outcome, estimating the causal effect. This method does not require individual-level data, making it suitable for analysis based on summary statistics.

To assess heterogeneity among SNPs in the model, this study conducts heterogeneity tests [14] . If significant heterogeneity is detected, it is advisable to focus more on the results of the IVW model [15] . Additionally, sensitivity analyses were conducted using the Leave-one-out method to examine the impact of individual SNPs on the model’s estimates.

All analyses are conducted using the TwoSampleMR package in R software (version 4.3.3). The significance level is set at α = 0.05 to ensure the reliability of the statistical results.

As shown in Figure 1 , the GWAS data for the alcohol consumption group yielded 6,483,125 SNPs. After screening with P < 5 × 10⁻⁸, 2147 SNPs remained. After reducing bias due to linkage disequilibrium, 8 SNPs remained. The POAG GWAS data revealed 5,961,428 SNPs. After integrating the data from both studies, 6 SNP datasets remained, as detailed in Table 1 .

Heterogeneity analysis was performed on these results, with a P-value of 0.44 (>0.05), indicating no significant heterogeneity and thus high reliability. Analysis and causal inference using these 6 SNP data through the three regression formulas showed that an increase of one standard deviation in alcohol abuse was associated with less than 1 increase in POAG risk, which was not statistically significant, as detailed in Table 2 . The pleiotropy analysis showed a P-value of 0.42, indicating no other unknown interfering factors. Scatter plots clearly show that both IVW and WME regression lines pass through the origin, also indicating no other unknown interfering factors. Scatter plot situations and stability tests are detailed in Figure 2 and Figure 3 , showing good stability and high reliability of the results.

The outcome heatmap situation is detailed in Figure 4 , where, except for rs1229984 which shows lower reliability, the remaining SNP sites are filled in blue, representing lower standard errors, meaning the corresponding outcome estimates are more precise and reliable. These points are all close to zero, indicating that alcohol abuse does not significantly increase the risk of developing POAG.

As shown in Figure 1 , the GWAS data for the smoking group comprised 1,157,911 SNPs. Following selection based on P < 5 × 10⁻⁸, 848 SNPs remained, with 306 SNPs left after reducing bias due to linkage disequilibrium. Reading POAG GWAS data retrieved 5,961,428 SNPs. After integrating both datasets, 253 SNPs remained, detailed in Table 3 .

Heterogeneity analysis of these results showed a P-value < 0.05, indicating significant heterogeneity; therefore, the focus was primarily on the results of the IVW regression model. Analysis and causal inference using these 253 SNP data through various regression formulas showed P-values all greater than 0.05, detailed in Table 4 . Given the considerable heterogeneity in this part of the results, consideration should be given to the results of the IVW regression model. An increase of one standard deviation in smoking was associated with an increased POAG risk of 0.94, but this was still not statistically significant. Bias conditions and stability tests are detailed in Figure 5 and Figure 6 , showing good stability and high reliability of the results. The heatmap of the analysis results, detailed in Figure 7 , shows that most SNP effect values are roughly symmetrically distributed around 0.02 and −0.02. Although the large presence of SNP outcome sites inevitably leads to high heterogeneity in this part, the heatmap reveals that

smoking is not a factor causing an increased risk of POAG.