Multiple sclerosis (MS) is an autoimmune demyelinating disease of the central nervous system. MS is associated with an increased risk of other autoimmune diseases such as inflammatory bowel disease, psoriasis, and autoimmune thyroiditis, but the risk of noninfectious uveitis (NIU) is especially pronounced. ^,, The prevalence of NIU in population-based MS studies varies widely from 0.4% to 28.6%, though ophthalmologic confirmation of uveitis diagnoses is variable and often lacking. Even at 1% to 2%, the prevalence of NIU is still about 10 times higher in people with MS compared to the general population. ^,, While intermediate uveitis is the most common form of uveitis in the MS population, any type of NIU may occur, and a history of uveitis is associated with greater MS disease activity. The pathogenesis of both NIU and MS are thought to involve a complex constellation of genetic and environmental risk factors, some of which overlap between the two conditions (eg, HLA-DR2 haplotype).

Over the last three decades, nearly 20 disease-modifying therapies (DMTs) have been developed to treat MS, with a commensurate number of clinical trials demonstrating their effectiveness in reducing MS relapses and preventing radiographic progression and disability. However, whether DMTs affect the risk of NIU remains unknown due to its lack of inclusion as a secondary endpoint in clinical trials, and existing evidence is limited to case reports and small case series. ^,, This lack of knowledge limits our ability to (1) consider uveitis history and risk factors when making personalized MS treatment decisions, and (2) explore whether treatments that were originally developed to treat MS could be repurposed to treat NIU in other patient populations. The objective of this study was to determine the comparative risk of developing NIU after DMT initiation for MS.

There were 48,221 DMT treatment episodes for 43,501 people with MS who met eligibility criteria ( Figure 1 ). The number of NIU outcomes and duration of follow-up for each DMT group are shown in Table . Baseline characteristics and standardized mean differences between treatment groups are presented in Supplemental Tables 2 to 6. All groups achieved excellent covariate balance after overlap weighting (defined as SMD <0.10).

Flowchart of inclusion and exclusion criteria.

The adjusted cumulative incidence of NIU is shown by DMT group in Figure 2 . People taking nucleic acid synthesis inhibitors or S1P modulators had an 84% lower hazard of developing incident NIU compared to those prescribed glatiramer acetate (adjusted hazard ratio [aHR] 0.16, 95% CI: 0.08-0.32). Interferons (aHR 0.68, 95% CI: 0.51-0.91) and fumarates (aHR 0.51, 95% CI: 0.41-0.79) were also associated with a lower hazard of NIU compared to glatiramer acetate ( Table ). Natalizumab (aHR 0.80, 95% CI: 0.56-1.13) and anti-CD20 monoclonal antibodies (aHR 0.85, 95% CI: 0.70-1.04) were not associated with a significant difference in NIU compared to glatiramer acetate. Using an alternative outcome definition of NIU diagnosis with corticosteroid treatment within 120 days, nucleic acid synthesis inhibitors/S1P modulators (aHR 0.16, 95% CI: 0.08-0.33) and fumarates (aHR 0.56, 95% CI: 0.38-0.82) remained similarly associated with a lower hazard of NIU compared to glatiramer acetate, but interferons did not (Supplemental Table 7). Adjusting for uveitis history rather than excluding people with a history of uveitis produced results that were similar to our primary analysis (Supplemental Table 8). The addition of a 60-day washout period after treatment switch also had no significant effect on our results (Supplemental Table 9).

Cumulative incidence of noninfectious uveitis by treatment group. Adjusted cumulative incidence of nucleic acid synthesis inhibitors/S1P modulators (A), anti-CD20 monoclonal antibodies (B), fumarates (C), natalizumab (D), and interferons (E) compared to glatiramer acetate. IFN = interferons.

In this study, we compared the risk of developing NIU after DMT initiation for MS. We found that several drug classes were associated with a lower incidence of NIU compared to glatiramer acetate, most notably fumarates and a combined class of nucleic acid synthesis inhibitors and S1P modulators. These findings suggest that certain DMTs are associated with a lower risk of uveitis, and if substantiated by additional research, this could have future treatment implications in both MS and other patient populations.

Insufficient sample size and a lower than expected event rate necessitated pooling nucleic acid synthesis inhibitors with S1P modulators, which does limit precision and raises the possibility of type I error. This combined medication class represents a heterogenous group of therapies with respect to mechanism of action and efficacy, and thus, we were unable in this exploratory work to determine whether both classes were similarly associated with uveitis or whether one class was primarily responsible for the observed reduction in NIU incidence. Nevertheless, there are plausible mechanistic reasons for both nucleic acid synthesis inhibitors and S1P modulators to reduce uveitis risk, and we believe this represents a promising signal warranting further investigation. Nucleic acid synthesis inhibitors reduce lymphocyte production and maturation, and other medications with a similar mechanism of action (eg, methotrexate) are currently used to treat NIU. While teriflunomide is considered one of the lower potency MS DMTs, reducing the annualized relapse rate by about 36%, cladribine is considered high potency, reducing relapses by about 60%. S1P modulators prevent the egress of lymphocytes from lymph nodes, impounding autoreactive lymphocytes away from their target tissue irrespective of the autoimmune process at play. In MS, they have been shown to reduce the annualized relapse rate by about 50% to 55%. While this mechanism has not specifically been studied in the context of uveitis, it is worth noting that one S1P modulator, ozanimod, is FDA approved for the treatment of ulcerative colitis, a condition that is also associated with an increased risk of NIU.

It is important to consider selection bias and residual confounding as potential explanations for the apparent risk reduction seen with nucleic acid synthesis inhibitors and S1P modulators. Macular edema in the absence of other signs of intraocular inflammation is a known side effect of S1P modulators, affecting <1% of patients, and patients must undergo routine ophthalmologic monitoring while taking these medications. In post hoc analyses of clinical trial data, history of uveitis and past ocular surgery have been identified as risk factors for macular edema associated with S1P modulator use. Therefore, it is possible that the patients who are selected for treatment with S1P modulators are biased toward a lower baseline risk of NIU. In a pooled analysis of clinical trial participants taking fingolimod, the prevalence of NIU was 0.59%. This is somewhat lower than the 1% to 2% uveitis prevalence reported in the general MS population, which suggests that uveitis history and risk factors could play a role in MS treatment selection. However, we excluded all patients with a prior history of NIU in our primary analysis, so the extent to which this affected our results is unclear. Similarly, enhanced ophthalmologic surveillance of patients on S1P modulators might have been expected to increase NIU detection, but this would have biased our results toward an increased rather than decreased NIU risk.

Other therapeutic classes were also associated with a reduced incidence of NIU. Fumarates stimulate the nuclear factor erythroid 2-related factor 2 (Nrf2) pathway, which is broadly involved in augmenting protective cellular responses to oxidative stress arising in the setting of inflammatory processes. While only approved in the United States to treat MS, in Europe, fumarates are also approved for psoriasis, which is also associated with an increased risk of uveitis. In a mouse model of human NIU, dimethyl fumarate has shown promise in attenuating the inflammatory process on histopathologic examination, but outside of our study, human data are lacking. Interferons are thought to work in part by modifying the cytokine milieu to take on a more anti-inflammatory profile (eg, increasing interleukin-10 and decreasing tumor necrosis factor-α [TNF-α]). This is noteworthy because TNF-α is known to play a prominent role in NIU, and the TNF-α antagonist adalimumab is the only FDA-approved systemic therapy for this condition. A small randomized controlled trial of interferon beta for intermediate uveitis with macular edema in MS demonstrated improvement in macular thickness compared to methotrexate at 3 months, but this has not been studied outside the context of MS, nor has it been previously studied in uveitis prevention prior to this study.

Since the approval of ocrelizumab for MS in 2017, anti-CD20 monoclonal antibodies have grown rapidly in popularity and are now the most commonly prescribed therapy for newly diagnosed MS patients. Ocrelizumab reduced relapse risk by 48% compared to interferons in its original phase III clinical trials, and data for ofatumumab and ublituximab (which followed in 2019 and 2022, respectively) are similar. Some long term, real-world effectiveness studies have since suggested up to a 95% reduction in relapse risk with anti-CD20 therapy. In a small retrospective case series of MS patients with treatment-resistant NIU, multiple measures of uveitis severity (eg, visual acuity, vitreous haze score, retinal vasculitis score, and central retinal thickness) improved after a median follow-up time of 28.5 months compared to before starting anti-CD20 therapy. Similar evidence exists for using anti-CD20 therapies to treat NIU in the setting of other autoimmune diseases such as juvenile idiopathic arthritis. Natalizumab reduces the annualized relapse risk by 68% and the risk of disability progression over 2 years by ∼48% compared to placebo. Concerns about progressive multifocal leukoencephalopathy have heavily curtailed natalizumab usage in MS and Crohn’s disease, but a more selective inhibitor (vedolizumab), is approved for inflammatory bowel disease and has been described as improving uveitis in a small case series.

Because of these distinct mechanisms of action, the associations we observed between different medication classes and NIU did not necessarily reflect their known relative effectiveness for MS. For example, interferons are similarly effective to glatiramer acetate in relapsing remitting MS, reducing attacks by about 30%, but we found that patients on interferons had a 33% lower hazard of developing NIU compared to glatiramer acetate. Similarly, cohort studies and network meta-analyses show that anti-CD20 therapies are superior to fumarates for MS, ^, yet our work suggests the potential for greater NIU risk reduction with fumarates. Natalizumab is one of the highest efficacy DMTs for relapsing forms of MS, but was not associated with a reduced risk of NIU compared to glatiramer acetate in our study. This may be due to its unique mechanism of action: natalizumab blocks the α4 subunit of the α4β1 and α4β7 integrins on the surface of lymphocytes that regulate their diapedesis from the circulation into tissues. α4β1 is most important for transmigration across the blood-brain barrier and α4β7 for transmigration across the gastrointestinal mucosa, but the relative importance of these particular integrins in other tissues (eg, the uvea) compared to the other ∼24 known human leukocyte integrins is less well understood. Thus, while the potential benefits of NIU prevention may be considered when selecting certain DMTs, this decision needs to be considered within the context of overall MS treatment effectiveness.

There are some additional limitations to our study. Patients were part of a private insurance network and thus are more likely to have experienced steady employment and higher socioeconomic status, potentially limiting generalizability. Administrative claims data do not allow independent verification of the accuracy of ICD codes with medical chart level data, but previous validation studies show that the accuracy of these codes for both MS and NIU is high. ^, We required initial NIU diagnoses to be confirmed with a second diagnosis code (or in our sensitivity analysis, a corticosteroid prescription) to maximize diagnostic accuracy and prevent outcome misclassification, but as a result, mild cases with spontaneous resolution may have been missed. The McDonald criteria for MS underwent several revisions between 2000 and 2022, so the actual clinical phenotypes of patients who otherwise received the same diagnosis codes may have changed over time. Finally, while we adjusted for multiple surrogate markers of MS severity and disease-related disability, data on MRI lesion burden and distribution (eg, spinal cord lesions, which portend more aggressive disease and greater disability), cerebrospinal fluid oligoclonal bands, and neurofilament light chain were unavailable. These are important determinants of MS treatment allocation regardless of disability status, but evidence for an association with NIU is mixed: in a large individual patient data meta-analysis, history of NIU was associated with a greater number of T2 lesions but fewer relapses in the preceding 2 years and no difference in the number of new enhancing lesions. Nevertheless, residual confounding is possible and could have influenced our findings.

In summary, we found in this exploratory work that people taking certain DMTs for MS may have a lower risk of NIU, especially those taking nucleic acid inhibitors, S1P modulators, or fumarates. If these findings are confirmed in additional studies of other cohorts (eg, inflammatory bowel disease, for which fumarates and S1P modulators are also prescribed), such medications could eventually have a role in treating MS patients who have a higher risk of NIU as well as treating NIU outside the context of MS. Before it was FDA-approved to treat uveitis, adalimumab had a long history of success in treating rheumatologic diseases associated with uveitis such as ankylosing spondylitis, inflammatory bowel disease, and psoriatic arthritis. Our work suggests that MS can serve as another “discovery cohort” for identifying novel treatments for NIU, and similar strategies may be applied to other patient populations too. ^,