Alcohol consumption and rheumatoid arthritis risk: A prospective cohort study with nonlinear Mendelian randomization analysis.

Abstract

Background: Previous epidemiological studies have shown a nonlinear relationship between alcohol consumption and rheumatoid arthritis (RA), though these findings may be biased by confounding factors or reverse causation. Additionally, the impact of sex on this association remains inconsistent. Therefore, we aim to investigate whether the causal link between alcohol consumption and RA risk is linear, nonlinear, or both.

Methods: Participants from the UK Biobank who provided detailed alcohol consumption information and complete covariate data were included in this study. Alcohol intake was quantified in units per week. We employed multivariable Cox models with restricted cubic splines for conventional analysis, and both linear and nonlinear Mendelian randomization (MR) analyses to assess causal relationships.

Results: Among the 316,717 participants, 3264 incident cases of RA were recorded during an average follow-up of 13.22 years. The Cox regression model suggested that the association between weekly alcohol consumption and RA incidence was an approximate U-shaped relationship, with the lowest risk at 21.95 units/week. Each unit increase in alcohol consumption was significantly associated with a lower risk of RA in women (HR: 0.991; 95% CI: 0.986, 0.996), but not in men (P for interaction <0.001). However, nonlinear MR did not detect a significant nonlinear correlation between alcohol consumption and RA risk, either overall (P for nonlinearity = 0.161) or within sex subgroups. The individual-level linear MR also indicated that genetically predicted alcohol consumption is not associated with RA risk.

Conclusions: The overall and sex-specific associations found in conventional epidemiological analyses were not supported by either linear or nonlinear MR analyses.