Association between patterns of circulating vitamin co-exposure and the prognosis of rheumatoid arthritis

This study aimed to examine the independent and joint associations of seven serum vitamins (vitamin A, B₆, B₉, B₁₂, C, D, and E) with prevalent rheumatoid arthritis (RA) and long-term all-cause mortality in a nationally representative sample of US adults, with rigorous adjustment for disease-modifying antirheumatic drugs (DMARDs) and glucocorticoids using linked prescription data.

We analyzed data from the National Health and Nutrition Examination Survey (NHANES) 2003-2006 cycles, linked to National Death Index (NDI) mortality records through December 31, 2019. Of 37,866 total participants, we included 14,487 adults aged ≥ 20 years who were non-pregnant and had complete interview and laboratory examination. This analytic sample comprised 682 RA cases (defined by affirmative responses to both MCQ160A and MCQ190), 1191 osteoarthritis (OA) patients, and 12,614 healthy controls (HC). Missing vitamin concentrations (6.1-7.1% across analytes) were handled via multiple imputation using the Multivariate Imputation by Chained Equations (MICE) algorithm (m = 5). To assess joint associations with prevalent RA, we applied survey-weighted logistic regression, weighted quantile sum (WQS) regression, Bayesian kernel machine regression (BKMR), and quantile g-computation (qgcomp). For survival analysis, Cox proportional hazards models were used to estimate hazard ratios (HRs) for all-cause mortality among RA patients, with DMARD and glucocorticoid use incorporated as time-varying covariates based on prescription history.

After full covariate and medication adjustment and application of NHANES sampling weights, no individual vitamin was significantly associated with prevalent RA risk. Similarly, neither the WQS index (OR [95% CI] = 1.032 [0.976-1.090], p = 0.262) nor the qgcomp joint effect estimate (OR [95% CI] = 1.033 [0.906-1.177], p = 0.626) reached statistical significance. In longitudinal analysis (median follow-up: 14.0 years), higher circulating levels of vitamin C (HR [95% CI] = 0.627 [0.473-0.831], p = 0.001), vitamin B₉ (HR [95% CI] = 0.975 [0.960-0.990], p = 0.001), vitamin D (HR [95% CI] = 0.988 [0.981-0.996], p = 0.002), and vitamin B₆ (HR [95% CI] = 0.997 [0.995-0.999], p = 0.006) were each independently associated with lower all-cause mortality risk among RA patients. Only 8.4% of RA patients received DMARD therapy during follow-up, nevertheless, HRs remained stable after incorporating medication exposure attenuation was minimal (< 5% change in point estimates) compared with unadjusted models. Kaplan-Meier survival curves further confirmed that RA patients in the highest tertile of vitamin D (T3) exhibited significantly better survival than those in the lowest tertile (T1) (log-rank p < 0.001).

Neither individual vitamins nor their complex mixtures showed statistically significant associations with prevalent RA after accounting for confounding, survey design, and pharmacologic treatment. In contrast, higher serum concentrations of vitamins C, D, B₉, and B₆ were robustly and independently associated with reduced long-term all-cause mortality in RA patients, even after strict adjustment for DMARDs, glucocorticoids, and major clinical and demographic covariates. These findings underscore the potential prognostic utility of routine serum vitamin profiling in the clinical management of RA. Key Points • There were no causal association was found between vitamins and RA prevalence, neither independently nor joint effects. • Among RA patients, elevated serum levels of vitamins C, B9, D, and B6 were each independently associated with reduced all-cause mortality, with vitamin C exhibiting the strongest and most precise protective effect (HR [95% CI] = 0.627 [0.482-0.815], p = 0.001). • After adjustment of DMARD and glucocorticoid use, the associations of vitamins C, B9, D, and B6 with reduced all-cause mortality in patients with rheumatoid arthritis remained statistically robust.