Perfluoroalkyl Substance (PFAS) Mixtures Drive Rheumatoid Arthritis Risk Through Immunosuppression: Integrating Epidemiology and Mechanistic Evidence

Abstract

Perfluoroalkyl substances (PFASs) possess immunosuppressive properties. However, their association with rheumatoid arthritis (RA) risk remains inconclusive across epidemiological studies. This study integrates population-based and mechanistic evidence to clarify the relationship between PFAS exposure and RA. We analyzed 8743 U.S. adults from the NHANES (2005-2018), assessing individual and mixed exposures to PFOA, PFOS, PFNA, and PFHxS using multivariable logistic regression, Bayesian kernel machine regression, quantile g-computation, and weighted quantile sum models. Network toxicology and molecular docking were utilized to identify core targets mediating immune disruption. The results showed that elevated PFOA (OR = 1.63, 95% CI: 1.41-1.89), PFOS (OR = 1.41, 1.25-1.58), and PFNA (OR = 1.40, 1.20-1.63) levels significantly increased RA risk. Mixture analyses indicated a positive joint effect (WQS OR = 1.06, 1.02-1.10; qgcomp OR = 1.26, 1.16-1.38), with PFOA as the primary contributor. Stratified analyses revealed stronger effects in females (PFOA Q4 OR = 3.75, 2.36-5.97) and older adults (≥60 years). Core targets included EGFR, SRC, TP53, and CTNNB1. PFAS mixtures increase RA risk, dominated by PFOA and modulated by sex/age. These findings help reconcile prior contradictions by identifying key molecular targets and vulnerable subpopulations, supporting regulatory attention to PFAS mixture exposure.

Keywords: Bayesian kernel machine regression (BKMR); NHANES; mixture exposure; molecular docking; perfluoroalkyl substances (PFAS); rheumatoid arthritis.

Figure 1

The Pearson correlation between PFAS metabolites after being log-transformed.

Figure 2

Odds ratios (95% CI) of RA associated with co-exposure to PFAS metabolites by WQS (A) and qgcomp (B) analyses in total population and subgroups. Models were adjusted for age, sex, race, education level, marital status, family PIR, BMI (kg/m²), smoking, drinking alcohol status, diabetes.

Figure 3

The joint effect of urinary PFAS metabolites on RA was estimated by BKMR models. (A) the total population; (B) 20 ≤ Age ≤ 39; (C) 40 ≤ Age ≤ 60; (D) Age ≥ 60.

Figure 4

The univariate expose–response function (95% CI) between PFAS metabolites and RA was estimated using the BKMR model. (A) the total population; (B) 20 ≤ Age ≤ 39; (C) 40 ≤ Age ≤ 60; (D) Age ≥ 60.

Figure 5

Network toxicology analysis. (A) Venn diagram of PFAS-related targets; (B) Venn diagram of RA-related targets; (C) Venn diagram of overlapping PFAS- and RA-related targets; (D) bar plot of Gene Ontology (GO) enrichment analysis; (E) bubble plot of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis; top 10 hub targets calculated via (F) betweenness, (G) bottleneck, (H) closeness, (I) degree, (J) maximum neighborhood component (MNC), (K) radiality, and (L) stress; (M) Venn diagram illustrating target overlaps identified by seven topological algorithms; (N) protein–protein interaction (PPI) network of the four core targets.

Figure 6

Molecular Docking. (A) PFOA-CTNNB1; (B) PFOA-EGFR; (C) PFOA-TP53; (D) PFOA-SRC; (E) PFOS-CTNNB1; (F) PFOS-EGFR; (G) PFOS-TP53; (H) PFOS-SRC; (I) PFNA-CTNNB1; (J) PFNA-EGFR; (K) PFNA-TP53; (L) PFNA-SRC.