Tumor Necrosis Factor-α Inhibition and the Unmasking of CNS Demyelination Across a Large U.S. Health Research Network

To evaluate the risk of demyelinating disease and optic neuritis among patients treated with TNF-α inhibitors using a large, nationwide electronic health record database. Retrospective, propensity score–matched cohort study. A total of 283,760 patients with newly initiated autoimmune disease (excluding CNS demyelination) on TNF-α inhibitors were matched 1:1 to 283,760 controls with similar diagnoses who did not receive TNF-α inhibitors. Electronic health records from the TriNetX network (70 U.S. healthcare organizations) were analyzed from 2010–2025. Demographics and comorbidities (including vitamin D deficiency, obesity, thyroid disease, mononucleosis, smoking) were balanced using propensity score matching. Time-to-event analysis was performed to estimate hazard ratios (HRs) for demyelinating disease and optic neuritis. Incidence and hazard ratios of demyelinating disease and optic neuritis at 1, 1–3, and 3–5 years after initiation of TNF-α inhibitor therapy. Across all TNF-α inhibitors, exposure was associated with a modest increase in the risk of demyelinating disease or optic neuritis at 1–3 years (HR 1.237, 95% CI 1.081–1.415 ) and 3–5 years (HR 1.201, 95% CI 1.003–1.438 ). However, the absolute incidence of these events remained notably low (all <1%). These signals were not uniform across the drug class. While an increased 5-year risk was observed with infliximab (0.531% vs 0.244% in controls; HR 1.983), other agents such as etanercept, certolizumab, and golimumab did not show statistically significant increases for the combined outcome. Furthermore, adalimumab showed a significant association specifically with optic neuritis (HR 1.278, 95% CI 1.013–1.613), but not with the combined outcome or demyelinating disease alone. TNF-α inhibitor exposure was associated with a small and statistically significant increase in the risk of demyelinating events, though the absolute risk remains very low. The variability in risk signals across different inhibitors and the low overall incidence suggest that while clinical awareness is appropriate, these findings should be viewed within the context of the established benefits of these therapies.