PFAS Exposure Linked to Rheumatoid Arthritis Risk

Key Takeaways

• In a bidirectional two-sample MR analysis using European-ancestry genetic datasets from FinnGen, UK Biobank, and an 8,299-person exposure GWAS, higher genetically predicted PFOA was significantly associated with higher RA risk.
• Genetically predicted PFOS showed a positive signal concentrated in seropositive RA phenotypes rather than across all RA definitions examined.
• Sensitivity analyses did not identify significant heterogeneity, directional pleiotropy, MR-PRESSO outliers, or single-variant dominance behind the reported PFAS-RA associations.
• Reverse MR did not show evidence that genetically predicted RA altered circulating PFOA or PFOS levels.

Per- and polyfluoroalkyl substances (PFAS) are persistent environmental exposures that may intersect with the genetic and environmental susceptibility profile of rheumatoid arthritis (RA), but observational studies have not resolved whether that relationship is causal. RA remains a clinically heterogeneous autoimmune disease in which environmental triggers are difficult to separate from confounding and reverse causation. Mendelian randomization offers a way to test whether inherited liability to higher circulating PFAS tracks with RA risk, and investigators used that approach to examine both directions of the association.

Investigators conducted a bidirectional two-sample Mendelian randomization (MR) analysis of genetically predicted circulating perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) in relation to rheumatoid arthritis (RA) risk. Exposure instruments came from a genome-wide association study (GWAS) of 1,091 blood metabolites and 309 metabolite ratios in 8,299 European individuals from a Canadian population-based study, and RA outcomes came from FinnGen and the United Kingdom (UK) Biobank across multiple phenotypes; FinnGen datasets ranged from 1,297 to 6,329 cases and 147,221 to 215,389 controls. In the methods section, the authors report that no SNPs met genome-wide significance, so they selected IVs at P<1×10^-5, clumped them at R2 0.001 over 10,000 kb, and excluded SNPs with F-statistics below 10. Inverse variance weighted (IVW) analysis was the primary method, with MR-Egger, weighted median, simple mode, weighted mode, maximum likelihood estimation, and robust adjusted profile score as supporting approaches, alongside Cochran Q, the MR-Egger intercept, Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO), and leave-one-out testing.

In the primary IVW analysis, the Mendelian randomization study in Medicine found that genetically predicted PFOA was associated with RA risk, with an odds ratio (OR) of 1.87, a 95% confidence interval (CI) of 1.13-3.11, and P=.015. A phenotype-specific PFOS signal was also reported for strict seropositive RA, with OR 1.14, 95% CI 1.00-1.29, P=.049, and additional seropositive PFOS associations were directionally similar. Alternative MR approaches were directionally consistent, sensitivity analyses did not identify significant heterogeneity, directional pleiotropy, MR-PRESSO outliers, or single-SNP dominance, and reverse MR did not show significant causal effects of genetically predicted RA on circulating PFOA or PFOS concentrations.

The analysis relied primarily on European-ancestry datasets from Finland, the UK, and Canada, which limits generalizability beyond those populations and makes the findings non-U.S.-specific. For North American clinicians, the results are best read as etiologic evidence rather than as a U.S. population estimate. The exposure GWAS was relatively small and required a relaxed SNP-selection threshold, which the authors identified as a potential source of weak instrument bias. Because the analysis used genetic proxies for PFAS exposure rather than measured real-world exposure dynamics, it cannot show what would happen after exposure reduction or define the underlying biological mechanism. The authors also noted that residual confounding remains possible despite the MR design.

According to the authors, these findings support a potential association between genetically predicted PFAS exposure and RA risk, with the clearest signal for PFOA and additional phenotype-specific PFOS associations in seropositive RA. They framed the results as a rationale for further mechanistic research, gene-environment investigation, and study of long-term health effects rather than as proof of a defined pathway or intervention benefit.

Clinician Questions

Which rheumatoid arthritis phenotypes showed the PFOS association in this Mendelian randomization analysis?

The PFOS signal was concentrated in seropositive rheumatoid arthritis phenotypes, including the strict seropositive definition, rather than being reported uniformly across all rheumatoid arthritis definitions assessed in the analysis.

Why does the SNP selection threshold matter when interpreting the PFAS-rheumatoid arthritis signal?

No SNPs for PFOA or PFOS met the usual genome-wide significance threshold, so investigators used a more lenient P<1×10^-5 threshold, clumped variants for independence, and excluded SNPs with F-statistics below 10. The authors identified that approach as a potential source of weak instrument bias, which tempers how definitively the PFAS-rheumatoid arthritis association can be interpreted.

What does reverse Mendelian randomization add to the interpretation of PFOA, PFOS, and rheumatoid arthritis risk?

Reverse MR tested the opposite direction by asking whether genetically predicted rheumatoid arthritis influenced circulating PFOA or PFOS concentrations, and it did not show significant causal effects. That finding supports the reported directionality of the PFAS-rheumatoid arthritis signal without establishing mechanism or showing that lowering exposure would change clinical risk.