In 109,565 UK Biobank adults followed a median 11.6 years, the highest Mediterranean diet quintile had ~32% fewer rheumatoid arthritis (RA) cases than the lowest, and a 66-metabolite plasma signature showed an even stronger gradient. Both the diet score and the signature remained independently related to lower RA risk after mutual adjustment, with linear dose–response trends and robust results across sensitivity and genetic-risk analyses. The metabolic signature mediated 22.4% of the diet–RA relationship, with lipid markers (e.g., omega-3 measures, DHA, degree of unsaturation) contributing most. Limitations included single-timepoint metabolomics, incomplete repeat measures, a predominantly White cohort, and the observational design.