Alcohol, Coffee, and Milk Intake in Relation to Epilepsy Risk

Alcohol, coffee and milk intakes have been explored in relation to epilepsy risk in observational studies; however, the results were not consistent. We performed a Mendelian randomisation (MR) study to evaluate the causality of these relationships. Genetic variants associated with alcohol, coffee and milk intake were adopted as instrumental variables. We obtained the summary data of epilepsy from the International League Against Epilepsy (ILAE) Consortium (15,212 cases and 29,677 controls) and FinnGen consortium (4588 cases and 144,780 controls). Genetically predicted alcohol intake was associated with a higher risk of epilepsy in the ILAE Consortium (odds ratio (OR): 1.22, 95% confidence intervals (CI): 1.02–1.45). The association in the FinnGen consortium remained consistent in direction. Combined analysis of ILAE and FinnGen databases further indicated that genetically predicted alcohol intake was associated with a higher risk of epilepsy (OR = 1.24; 95% CI, 1.06–1.47, p = 0.009). Genetically predicted coffee intake was not related to epilepsy risk, while higher genetically predicted milk intake was related to a lower risk of epilepsy (OR = 0.957; 95% CI, 0.917–0.999, p = 0.044). Our results suggest a detrimental effect of alcohol intake on the risk of epilepsy, while milk intake might be associated with a decreased risk of epilepsy.


Introduction
It is estimated that epilepsy affects over 70 million people worldwide [1], which brings a huge economic and social burden. Previous studies have been reported regarding the association between alcohol intake and epilepsy risk. However, these results are not consistent [2][3][4]. The effect of other modifiable lifestyle behaviors, such as coffee and milk intake, in epilepsy, has been scarcely explored [5,6]. Most notably, these results are at risk of reverse causation and confounding bias. Therefore, successful interventions targeting modifiable lifestyles to prevent epilepsy development are of great importance to lowering the disease burden. Moreover, alcohol, coffee and milk intake are changeable lifestyle behaviors; determining their causal relationships with epilepsy could reveal the potential pathogenesis of epilepsy and are of great significance for formulating efficient prevention measures.
Mendelian randomization (MR) adopts genetic variance as an instrumental variable to mimic the effect of environmental exposures, such as alcohol, coffee and milk intake, to explore causality [7]. Since the single nucleotide polymorphisms (SNPs) one is born with are constant through a lifetime, MR could decrease reverse causation bias, suggesting that the outcome (epilepsy) could not modify a person's genetic predisposition for the exposure. Moreover, SNPs are generally unrelated to other confounders, since SNPs are randomly assorted during conception. Therefore, we performed an MR study to evaluate the associations of alcohol, coffee, and milk intake with epilepsy risk.

SNP Selection
Instrumental variables of alcohol intake were obtained from the GSCAN (941,280 Europeans) [8], which identified 99 alcohol intake related-SNPs (drinks per week, p < 5 × 10 −8 ). Moreover, the 99 alcohol intake related-SNPs were clumped for independence (r 2 < 0.01; region size, 10 Mb) based on the Europeans data from the 1000 Genomes Project [9]. As a result, 84 independent SNPs were adopted as instrumental variables for alcohol intake (Supplementary  Table S1).
For milk intake, rs4988235 was adopted as an instrumental variable. SNP rs4988235 is strongly related to milk intake in European individuals, and an additional T-allele of rs4988235 was associated with 0.58 (95% CI: 0.49-0.68) glasses/week increase in milk intake among a Danish cohort (p = 9 × 10 −36 ) [11].
When these instrumental variables for alcohol, coffee and milk intake were not available in the epilepsy dataset, proxy SNPs in linkage disequilibrium (LD, r 2 ≥ 0.8) were adopted through LDlink (https://ldlink.nci.nih.gov/, accessed on 1 June 2021).

Outcome Data Sources
Genetic results for overall epilepsy were obtained from a large genome-wide analysis in the International League Against Epilepsy (ILAE) consortium (15,212 cases and 29,677 controls, Table 1) [12]. Details on cohorts and phenotype definition, study design, genotyping and quality control were described in the original study [12]. For each variant, the beta value and corresponding standard error were calculated with the approach described by Zhu et al. [13]. Moreover, we additionally adopted data from the FinnGen consortium (http://r4.finngen. fi/, accessed on 1 June 2021) to further validate the results for these exposures and epilepsy. The FinnGen Data Freeze 4 contains 4588 epilepsy cases and 144,780 controls (Table 1).
Our study was based on publicly available data only. Ethical approval for each of the GWASs could be found in the original publications.

Statistical Analyses
We applied the random-effects inverse-variance weighted (IVW) approach as the main analysis. Results from the ILAE and FinnGen consortium were combined with the fixed-effects meta-analysis. We adopted the weighted median as sensitivity analysis [14]. We also conducted MR-Egger regression to evaluate the directional pleiotropy [14]. Finally, the MR-PRESSO method was applied to identify potential outliers [14].
For alcohol drinking, the odds ratios (ORs) of epilepsy were scaled to a one standard deviation increase in log-transformed drinks per week. For coffee, the ORs of epilepsy were scaled to per 50% increase in coffee intake. For milk, the OR of epilepsy was calculated per additional T-allele of rs4988235. Association with a p-value less than 0.017 (0.05/3 exposures) was considered as significant, and an association with a p-value between 0.017 and 0.05 was considered as suggestive evidence of association. We used MendelianRandomization [15] and MR-PRESSO [16] packages in R software to conduct the analyses.

Selection of Genetic Variants
Of the 84 independent SNPs for alcohol intake, 24 SNPs were unavailable in the ILAE database. Suitable proxy SNP (r 2 ≥ 0.8 with the specified SNP) was available for 4 SNPs. As a result, 64 SNPs were available in the ILAE database. Moreover, the MR-PRESSO method identified two potential outliers (rs62044525 and rs10506274), which were excluded from subsequent analyses. Therefore, 62 SNPs were used in the ILAE database. In addition, 4 of the 84 SNPs were unavailable in the FinnGen consortium database. Suitable proxy SNP (r 2 ≥ 0.8) was available for 3 SNPs; therefore, 83 SNPs were used in the FinnGen consortium database.
Of the 12 independent SNPs for coffee intake, 7 SNPs were unavailable in the ILAE database. Suitable proxy SNP (r 2 ≥ 0.8 with the specified SNP) was available for 2 SNPs. As a result, 7 SNPs were available in the ILAE database. All the 12 SNPs were available in the FinnGen consortium database. For milk intake, rs4988235 was available in the FinnGen consortium. However, this genetic variant was unavailable in the ILAE database, and no suitable proxy variant was available at an LD r 2 ≥ 0.8. Thus, the association between milk intake and epilepsy was only analyzed in the FinnGen consortium.
Genetically predicted milk intake showed a suggestive association with a lower risk of epilepsy in the FinnGen consortium. The OR was 0.957 (0.917-0.999, p = 0.044, Figure 2B) for each additional milk intake increasing allele (T).

Discussion
In our study, we explored the causal association between alcohol, coffee and milk intake and epilepsy risk using the MR design. We found that genetically predicted alcohol intake was associated with a higher risk of epilepsy, while genetically predicted milk intake showed a suggestive association with a lower risk of epilepsy in the FinnGen consortium.
A recent MR study was also suggestive of a detrimental effect of genetically predicted alcohol consumption on the risk of epilepsy (OR, 1.54; 95% CI, 0.99-2.41; p = 0.058) [17]. The insignificant association in this previous study is likely to be caused by an inadequate power. A previous observational study showed that alcohol intake was associated with a higher risk of epilepsy in a dose-response manner [3]. Karhunen et al. found that chronic alcohol intake could damage the cerebellum and lead to a significant loss of Purkinje cells [18]. Moreover, people with alcohol dependence have a higher incidence of head trauma and brain injury due to traffic accidents, falls, or attacks [19]. Epilepsy is a severe nervous system disorder worldwide. Thus, it is very important to formulate efficient prevention measures for epilepsy. According to the result of our MR study, alcohol intake should be limited for those at high risk of epilepsy.
Very few investigations have reported the association between coffee/caffeine intake and epilepsy risk [4]. A previous study showed that long-term caffeine intake was not related to the risk of epilepsy [4]. Our results were consistent with the conclusion from the previous study [4].
Evidence on the association between milk intake and epilepsy is also limited. The potential protective effect of milk on epilepsy may be mediated through calcium and vitamin D. In a Portuguese epilepsy cohort, more than half of the epilepsy patients showed vitamin D deficiency [20]. In a south Indian study, the calcium intake of epilepsy patients was also far below the recommended level [21]. In addition to calcium and vitamin D, other nutrients in milk may also have an effect on epilepsy development, which needs further research.

Strengths and Limitations
Compared with conventional observational studies, the present MR study adopted summary data from GWASs with large sample sizes, which enabled us to draw clear conclusions and establish a causal relationship between these exposures and epilepsy. In addition, sensitivity analyses yielded similar results, which validate the robustness of the results. Moreover, we conducted the MR analyses in two large databases (the ILAE and FinnGen consortiums) and obtained consistent results, which further confirmed the association.
Our study also has some shortcomings. First, we cannot evaluate the nonlinear association between these exposures and the risk of epilepsy, owing to the linear effect assumption of the MR analyses. Second, our study is mainly based on Europeans. Our study may not have the same results in other populations with equal or greater intakes of alcohol, coffee or milk. Third, the effect of milk intake on epilepsy was only available in the FinnGen consortium. Thus, whether milk intake influences the risk of epilepsy warrants further research. Fourth, the associations of other diet factors with the risk of epilepsy were not investigated in our study, which is worth further study.

Conclusions
The present study provides evidence that a genetically predicted higher alcohol consumption was associated with an increased epilepsy risk. Reducing alcohol intake should be regarded as an important prevention strategy for epilepsy.