In a prospective cohort of 109,565 UK Biobank participants, patients in the highest quintile of the Mediterranean diet score had a 32% lower incidence of rheumatoid arthritis compared with the lowest quintile, with a dose–response pattern (Q5 vs Q1). The 66-metabolite plasma signature of relative lipoprotein lipid concentrations, fatty acids, amino acids, lipoprotein subclasses, and other measures showed a stronger gradient (Q5 vs Q1).
In mutually adjusted models, the Mediterranean diet (MED) diet score and the metabolic signature were each independently related to lower rheumatoid arthritis risk. Comparing the 90th with the 10th percentile, the hazard ratio was 0.73 for the MED diet score and 0.60 for the metabolic signature. Linear dose–response relationships were supported by restricted cubic splines. In analyses that jointly considered genetic susceptibility, patients with a high polygenic risk score (PRS) but the highest MED diet quintile had a risk that was similar to those with a medium PRS and the lowest MED diet quintile. Further, the study researchers observed a comparable pattern for the metabolic signature, where a high-PRS group with a favorable signature had similar or slightly lower risk than a medium-PRS group with an unfavorable signature. “Our study showed that the MED-related metabolic signature effectively captures the systemic biological response to dietary patterns,” wrote lead authors Xin Song and Xiaofeng Ma, of the Department of Epidemiology and Biostatistics at the West China School of Public Health and West China Fourth Hospital, Sichuan University in Chengdu, Sichuan, China. Mediation analysis estimated that the metabolic signature explained 22.4% of the diet–RA association. Twenty-two individual metabolites significantly mediated the diet–RA association (each explaining about 1% to about 9%). The largest proportions represented degree of unsaturation, ratio of omega-3 fatty acids to total fatty acids, docosahexaenoic acid (DHA), and the ratio of DHA to total fatty acids.
Participants were aged 37 to 73 years and completed 24-hour Oxford WebQ dietary recalls, provided plasma for metabolomics, and had genotyping. The MED diet score summarized recall data. Plasma biomarkers were quantified by nuclear magnetic resonance spectroscopy (Nightingale Health). Of 251 biomarkers screened, elastic-net regression yielded a MED diet–related metabolic signature of 66 metabolites: relative lipoprotein lipid concentrations (n = 31), fatty acids (n = 8), amino acids (n = 6), lipoprotein subclasses (n = 5), and other measures (n = 16). The signature correlated with the MED diet score across training, testing, combined, and repeat-assessment datasets (r ≈ 0.27–0.29). Incident RA was ascertained from linked health records. Cox proportional hazards models adjusted for demographics, socioeconomic and lifestyle factors, sleep duration, energy intake, fasting duration, body mass index, comorbidities, genotyping batch, and genetic principal components. Models included the diet score and the signature together to assess independence. Sensitivity analyses excluded early follow-up, additional adjustment for nonsteroidal anti-inflammatory drug use, other autoimmune diseases, and waist circumference; competing-risk models; and restrictions to complete cases and to participants with multiple dietary recalls. In subgroup analyses, inverse associations were more pronounced among patients who reported moderate-to-vigorous physical activity, healthy sleep duration of 7 to 8 hours, and body mass index of less than 30 kg/m².
In terms of limitations, diet and metabolite measurements were not repeated for all participants; therefore, single-timepoint metabolomics may not capture temporal variability. The NMR panel did not include the full metabolome. The cohort was predominantly White, which may limit generalizability. The observational design precluded causal inference, and mediation results were constrained by the cross-sectional timing of the mediator relative to diet.
Full disclosures can be found in the published study.
