Risk of newly developed atrial fibrillation by alcohol consumption differs according to genetic predisposition to alcohol metabolism: a large-scale cohort study with UK Biobank

This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Institutional Review Board (IRB No. E-2203-005-1302). The need for informed consent was waived as anonymized data were used. For reasonable requests, data are available through approval and oversight by the UK Biobank. The UK Biobank has approval from the North West Multi-centre Research Ethics Committee as a Research Tissue Bank approval.

The UK Biobank is a large-scale prospective cohort study of approximately 500,000 volunteers enrolled from 2006 to 2010 in the UK. The detailed design and preliminary results have been published elsewhere [23]. In brief, the UK Biobank database includes individual demographic information, diagnostic history, and genomic data. All participants signed a written informed consent form, and the UK Biobank received ethical approval from the National Health Service Research Ethics Service (reference 11/NW/0382).

The study design is illustrated in Fig. 1. In total, 502,392 individuals enrolled between 2006 and 2010 were identified. Subjects with a history of AF before enrollment (n = 3650) were excluded. Moreover, individuals without valid genetic data (n = 22,187) and those without valid answers to alcohol consumption questionnaires (n = 1061) were also excluded. Among these, we conducted this study among individuals in the white British population who self-reported as white British and were verified through principal component analysis of genetic ancestry. A final total of 399,329 individuals were included in this study and followed up until 2021. The index date was the baseline visit to the UK Biobank.

Using the UK Biobank database, the average alcohol intake per day (g/day) was calculated to evaluate alcohol consumption, and the participants were subsequently categorized into non-drinkers, mild-to-moderate drinkers (<30 g/day), and heavy drinkers (≥30 g/day).

Participants in the UK Biobank were genotyped using the Affymetrix UK Biobank Lung Exome Evaluation (UK BiLEVE) Axiom array or the Affymetrix UK Biobank Axiom array. As previously reported, the UK Biobank team performed central quality control and imputation [24]. To define genetic predisposition to alcohol metabolism, we calculated polygenic risk score (PRS) as the sum of the single nucleotide polymorphism (SNP)’s allele dosages of risk-increasing alleles weighted by their reported log odds ratios. SNPs associated with alcohol metabolism were identified from an existing large genome-wide association study from European ancestry (Supplemental Table 1) [19], which has been explored in other studies [25]. To compare effect sizes corresponding to the continuous PRS, we transformed each PRS of the corresponding analytical data set to the standard normal distribution (Supplemental Fig. 1). According to the PRS, we stratified subjects into three groups based on tertiles: the low PRS tertile group implies a genetic predisposition of slower alcohol metabolism, while the high PRS tertile group implies a genetic predisposition of faster alcohol metabolism.

Demographic data, anthropometric data, and medical history of hypertension, diabetes mellitus, myocardial infarction, dyslipidemia, chronic kidney disease, heart failure, and stroke were obtained. In addition, data on smoking history were collected. Smoking history was assessed using standardized questionnaires. Detailed definitions of comorbidities are provided in Supplemental Table 2.

Newly diagnosed AF was defined as the study endpoint. The endpoint was defined based on ICD-10 codes (I48). The follow-up duration was defined as the interval between the index date and the occurrence of incident AF.

Data are presented as numbers and frequencies for categorical variables and mean ± standard deviation or median with interquartile range for continuous variables. For categorical variables, the chi-square test or Fisher’s exact test was used as appropriate. One-way analysis of variance was used to analyze continuous variables between two or more groups. The annual event incidence rate (aIR) was calculated as the number of events per 1000 person-years (PY). Multivariate Cox proportional hazard regression models were used to estimate hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for the associations between genetic predisposition to alcohol metabolism, real alcohol consumption habits, and the risk of incident AF. Variables that initially achieved P < 0.2 in the univariate Cox regression analysis or those with clinical relevance in terms of AF risks were included in a multivariable model (Supplemental Table 3). The multivariable models were adjusted for covariates, including age, sex, previous history of hypertension, diabetes mellitus, myocardial infarction, dyslipidemia, chronic kidney disease, heart failure, stroke, and PRS tertile for alcohol metabolism or alcohol consumption habit, respectively. Subgroup analyses were conducted according to the PRS tertiles for alcohol metabolism and alcohol consumption habits using Cox models. The joint association was demonstrated, with individuals who had heavy drinking habits and low PRS tertile serving as the reference group. We further investigated whether the effects of genetic predisposition to alcohol metabolism could be associated with the risks of incident AF across varying levels of alcohol consumption, using structural equation model (SEM) analysis. With SEM, we aimed to explore the complex association between genetic predisposition, alcohol consumption, and risks of incident AF. A two-sided P value of <0.05 was considered statistically significant. Statistical analyses were performed using R programming version 4.2.1 (The R Foundation for Statistical Computing, Vienna, Austria). We manipulated the imputed genotype data and calculated the individual-level polygenic risk score using PLINK version 1.90 (https://​www.​cog-genomics.​org/​plink/​) [26].

In total, 399,329 individuals without previous history of AF (mean age 56.8 ± 8.0 years; men, 181,981 [45.6%]) were analyzed in this study. Hypertension was diagnosed in 104,879 individuals (26.3%), diabetes mellitus in 16,148 individuals (4.0%), myocardial infarction in 21,369 individuals (5.4%), dyslipidemia in 47,650 individuals (11.9%), chronic kidney disease in 16,695 individuals (4.2%), heart failure in 12,550 individuals (3.1%), and stroke in 8455 individuals (2.1%). More than half of the included participants were never smokers (54.7%), and approximately one-third were ex-smokers (35.0%).

Based on alcohol consumption habits, individuals were classified into three groups: non-drinkers, mild-to-moderate drinkers, and heavy drinkers. The baseline characteristics of each group are presented in Table 1. Heavy drinkers showed male preponderance, had a more frequent history of hypertension and dyslipidemia and substantially higher rates of smoking than non-drinkers and mild-to-moderate drinkers. They also showed a genetic predisposition to higher alcohol metabolism, as calculated by the PRS. In contrast, non-drinkers seemed to have a greater history of diabetes mellitus than other groups. The baseline characteristics classified by a genetic predisposition to alcohol metabolism according to PRS are shown in Table 2. Individuals with a higher PRS for alcohol metabolism tended to consume more alcohol.

During a median follow-up of 12.2 years (interquartile range, 11.5–12.8), there were 19,237 new cases of AF (4.0%). As shown in Table 3, the aIRs of AF were 5.50, 4.73, and 6.90 per 1000 PY for non-drinkers, mild-to-moderate drinkers, and heavy drinkers, respectively. In the univariate analysis, mild-to-moderate drinkers were associated with a lower risk of AF (HR 0.89, 95% CI 0.86–0.92), but heavy drinkers were associated with an increased risk of AF (HR 1.12, 95% CI 1.08–1.16). After adjusting for covariates, mild-to-moderate drinkers showed a decreased risk of AF (HR 0.96, 95% CI 0.92–0.99), while heavy drinkers consistently showed a higher risk of AF (HR 1.06, 95% CI 1.02–1.10). When analyzed as a continuous variable, one alcohol unit increment per day was significantly related to an increased risk of AF (HR 1.01, 95% CI 1.01–1.02).

When individuals were stratified according to PRS tertiles, the aIRs of AF were 5.26, 5.45, and 5.42 per 1000 PY for the low, middle, and high tertiles, respectively (Table 3). In the univariate analysis, the middle tertile group (HR 1.01, 95% CI 0.98–1.05) and high tertile group (HR 1.00, 95% CI 0.97–1.04) had similar risks of incident AF compared to the low tertile group. Similarly, the risk of incident AF was consistently equivalent in the middle tertile group (HR 1.01, 95% CI 0.97–1.04) and high tertile group (HR 0.99, 95% CI 0.96–1.03) compared to the low tertile group in the multivariate analysis.

To determine whether the predicted risks of incident AF, based on real alcohol consumption habits, could be modified by a genetic predisposition to alcohol metabolism, we performed a subgroup analysis according to PRS for alcohol metabolism (Fig. 2). In the low PRS tertile group, mild-to-moderate drinkers showed similar risks of AF (HR 0.97, 95% CI 0.92–1.03), whereas heavy drinkers showed a higher AF (HR 1.10, 95% CI 1.02–1.18) than non-drinkers. In contrast, in the middle PRS tertile group, mild-to-moderate drinkers and heavy drinkers showed equivalent risks of AF (HR 0.95, 95% CI 0.90–1.01 and HR 1.06, 95% CI 0.99–1.14), compared to non-drinkers. Similar findings were observed in the high PRS tertile group; mild-to-moderate drinkers and heavy drinkers showed a similar risk of AF (HR 0.94, 95% CI 0.89–1.00 and HR 1.02, 95% CI 0.95–1.10) compared to non-drinkers. The subgroup analysis according to alcohol consumption habits is presented in Supplemental Table 4. When we analyzed joint association across genetic predisposition to alcohol metabolism and alcohol consumption habits, individuals in the low PRS tertile group appeared to be more susceptible to incident AF among drinkers (Fig. 3). In SEM analysis, we found associations between genetic predisposition to alcohol metabolism, alcohol consumption habits, and the risk of AF (Supplemental Table 5) after assessment of the goodness-of-fit parameters indicated an adequate fir for this model (Supplemental Table 6).