Neuroinflammation Associated With Tumor Necrosis Factor-α Inhibitor Exposure

Abstract

Objective

To identify the frequency and clinical spectrum of neuroinflammation associated with exposure to tumor necrosis factor-α inhibitors (TNFi).

Methods

We performed a single-system retrospective cohort study. Adults in the Yale New Haven Health System with documented use of any US Food and Drug Administration-approved TNFi between 2007 and 2017 were identified via automated review of the electronic medical record. Those who also had brain MRIs were identified and categorized as either TNFi exposed or unexposed. Individuals with MRI findings concerning for neuroinflammation were identified, and detailed chart reviews were performed.

Results

A total of 4,391 patients received TNFi, and 547 also had brain MRI. After exclusion criteria were applied, 375 MRIs occurred after TNFi exposure, and 132 MRIs occurred before TNFi. MRIs were normal for 20.8% of exposed patients. The most common abnormal finding was nonspecific, punctate T2 hyperintensities. Seventeen cases (4.5%) among the exposed cohort had findings consistent with neuroinflammation, of which 58.8% required TNFi discontinuation and additional immunotherapy, whereas an additional 23.5% discontinued TNFi alone. After 3 years, 70.6% had stable MRI findings, whereas 11.8% demonstrated progression. The 10-year period prevalence of neuroinflammation in all subjects exposed to TNFi was 0.4%.

Conclusions

Neuroinflammatory phenomena following TNFi are a common concern for those treating patients with autoimmune disease. This is a large-scale study identifying the epidemiology surrounding this phenomenon. TNFi-associated inflammation was a rare outcome in our cohort. Most treated patients had either normal or nonspecific MRI findings. Further risk stratification parameters need to be identified.

Tumor necrosis factor-α inhibitors (TNFi) are frequently used to treat a range of autoimmune conditions, including inflammatory bowel diseases (IBD), psoriasis, and inflammatory arthritis. TNF mediates inflammation through activation of a number of mechanisms, including proinflammatory cytokines, B and T lymphocytes, and macrophages, along with limiting regulatory T-cell function.¹ However, TNF blockade has been associated with worsening of multiple sclerosis (MS) and emergence of secondary demyelinating disease in case reports.^2–5 Early trials of TNF blockade to treat MS were halted due to worsening of disease in what appeared to be a dose-dependent manner.^6–8 Despite small numbers of cases reported, many providers are reluctant to use TNFi due to concern for iatrogenic demyelinating disease.^9,10 There are no practice guidelines or algorithms to help risk stratify patients before initiation of these common medications. With the prevalence of MS rising, understanding iatrogenic causes of neuroinflammation is paramount.¹¹

To evaluate the scope of iatrogenic neuroinflammation arising in the context of TNFi exposure, we evaluated neurologic outcomes for all patients exposed to TNFi within the Yale New Haven Health System (YNHHS) over a 10-year period. This represents one of the first large-scale studies systematically evaluating the prevalence of neurologic complications in this population. Our results will help guide clinical practice.

Methods

We performed a single-system retrospective cohort study. Adult patients (aged ≥18 years) in the YNHHS with documented use of US Food and Drug Administration-approved TNFi between 2007 and 2017 in the active medication list were identified via automated review of the electronic medical record. An additional search for TNFi was performed using text mining; charts identified by this methodology were manually reviewed by investigators to confirm exposure.

The subset of patients who had undergone MRI brain imaging was identified. The presence of a brain MRI was used as a surrogate to identify individuals who may have experienced an adverse neurologic event. The exposure group consisted of subjects who had TNFi exposure before brain MRI. Date of TNFi exposure was obtained from documentation of drug administration in the medical record or documentation by providers. In cases in which a single date could not be identified, exposure date was considered as the nearest known month or year of first exposure.

Because systemic autoimmune diseases sometimes cause neurologic pathology, a subset of patients with matching medical diagnoses (e.g., inflammatory arthritis, IBD, and lupus) who were unexposed to TNFi at the time of brain MRI were identified as controls.

For the exposed group, detailed chart reviews were performed, including demographics, type and duration of TNFi use, social and medical history, family history, indications for MRI, results of MRI, and whether the immunomodulatory regimen was changed in response to MRI results. Patients with radiologic evidence of neuroinflammation of the CNS—brain, cranial nerves, and meninges—were considered potential cases. All potential cases were adjudicated by a neuroimmunology subspecialist. In questionable cases, a second neuroimmunology subspecialist also reviewed the chart, and case status was assigned via consensus. Comparable chart reviews were performed for the unexposed cohort, including results of MRI.

For both exposed and unexposed groups, patients with preexisting diagnoses of neurosarcoidosis and MS were excluded from data analysis. Data entry was completed using a standard data dictionary, developed in advance of data collection. Specifically, MRI results were scored according to predefined criteria based on details of the radiology report and with direct review of imaging in ambiguous cases (appendix e-1, links.lww.com/CPJ/A231).

Standard Protocol Approvals, Registrations, and Patient Consents

Approval from an ethical standards committee to conduct this study was received from the Yale University Institutional Review Board.

Data Availability

Anonymized data will be shared by request from any qualified investigator.

Results

During the 10-year period, a total of 4,391 subjects were identified to have had TNFi exposure, 547 of whom had a brain MRI. After exclusion criteria were applied, 375 patients had been exposed to TNFi before MRI and 132 patients had no TNFi exposure at the time of MRI. Twenty-one of the unexposed cohort received TNFi at a later date and are represented in the exposed cohort (figure 1).

Figure 1. Patient Selection Flow Diagram.

Demographics between the TNFi-exposed and unexposed groups were similar—the majority were female (65.9% and 68.9%), White/Caucasian (80.3% and 72.7%), and never smokers (48.5% and 50.8%). Within the exposed cohort, the majority had a primary autoimmune disease (87.2%) and had been exposed to 1 TNFi (79.5%). The remaining 12.8% had received TNFi for other indications, most commonly as a treatment for immunotoxicity related to cancer immunotherapy. The most common autoimmune diagnosis in both exposed and unexposed groups was inflammatory arthritis (46.4% and 40.9%), followed by IBD (22.9% and 25.0%) (table 1).

Table 1.

Demographics of TNFi-Unexposed and TNFi-Exposed Cohorts

Within the exposed cohort, MRIs were often ordered for nonlocalizing complaints (n = 145, 38.7%) such as malaise, headache, dizziness, or generalized weakness. Localizing complaints, such as focal weakness/numbness, trigeminal neuralgia, or optic neuritis, occurred in 126 (33.6%) patients. Fifty-seven (15.2%) patients had surveillance MRIs, typically for cancers or postoperative monitoring, and the remainder of scans (n = 47, 12.5%) were performed to evaluate encephalopathy or seizures. The majority of MRIs were either normal (n = 78, 20.8%) or demonstrated only a few nonspecific, punctate T2 hyperintensities (n = 128, 34.1%). A few scans demonstrated many nonspecific, punctate T2 hyperintensities (n = 25, 6.7%) or confluent white matter changes (n = 14, 3.7%). Twenty-two (5.9%) MRIs had findings requiring further review by investigators, after which 17 (4.5%) were deemed to have neuroinflammatory changes, discussed in detail below. Some MRIs (n = 102, 27.2%) were read as abnormal due to findings unrelated to the outcome of interest, such as nonpathologic anatomic variants, surgical changes, acute and chronic ischemic and hemorrhagic strokes, and mass lesions (figure 2).

Figure 2. MRI Results of TNFi-Unexposed Cohort and TNFi-Exposed Cohort (Including Cases).

TNFi = tumor necrosis factor-α inhibitors.

Most TNFi-unexposed brain MRIs were also normal (n = 45, 34.1%) or demonstrated only a few nonspecific, punctate T2 hyperintensities (n = 46, 34.8%). Findings concerning for neuroinflammation were present in 5 (3.8%) unexposed subjects: posterior reversible encephalopathy syndrome, exacerbation of established neuro-Bechet disease, Bell palsy (commonly known as Bell's palsy), and idiopathic hypertrophic pachymeningitis.

TNFi-Associated Neuroinflammation

As stated, 22 TNFi-exposed subjects had brain MRI findings initially concerning for neuroinflammation; however, manual chart reviews led to the exclusion of 5 subjects. Three subjects were excluded due to having alternative identified etiologies for the brain MRI changes: CNS toxoplasmosis on brain biopsy and 2 primary brain tumors. A fourth subject was excluded due to a confounding diagnosis of neuro-Behcet disease. A fifth subject was excluded due to insignificant MRI findings. With 17 remaining cases of idiopathic neuroinflammation in the TNFi-exposed group, there was a 10-year period prevalence of 0.4%.

Similar to the overall cohort, cases tended to be female (76.5%), of White/Caucasian ethnicity (82.4%), and middle aged (mean age 48 years). Smoking history and family history of autoimmunity were more common among cases than in the overall cohort (table 1). All cases were receiving TNFi to treat an underlying autoimmune disease. The most common underlying autoimmune diseases were IBD (n = 7, 41.2%), followed by inflammatory arthritis (n = 5, 29.4%) and psoriasis (n = 3, 17.6%). Five subjects (29.4%) had more than 1 autoimmune diagnosis. Of interest, 5 subjects (29.4%) had a documented family history of autoimmune disease, with the most common diagnoses among relatives being IBD (n = 4, 80.0%) and inflammatory arthritis (n = 1, 20.0%).

The calculated mean time from initial TNFi exposure to first MRI findings of neuroinflammation was 6.5 years (±5.3 years). This span ranged from less than 1 year up to 17 years, the latter noted in a patient with years of continuous exposure to TNF blockade. Some subjects received only a few doses of TNFi, whereas others had many years of exposure, but this number does not take into account timing of the first neurologic complaint, merely the date of first pathologic brain MRI. Cerebrospinal fluid was obtained in 8 patients, notable for lymphocytic pleocytosis (n = 4, 50.0%) and positive oligoclonal bands (n = 2, 25.0%).

On being diagnosed with neuroinflammatory disease, the majority (n = 10, 58.8%) had the TNFi discontinued and required additional immunotherapy, such as steroids or other immunomodulatory therapy to treat the acute neuroinflammation. Four individuals (23.5%) had the TNFi discontinued without additional immunotherapy being needed. Two individuals (11.8%) had had the TNFi stopped before the first pathologic MRI on clinical grounds, specifically due to lack of treatment efficacy. TNFi was continued without changes for 1 individual. The majority had stable MRI findings at 3 years (n = 12, 70.6%), whereas a smaller subset had either progression (n = 2, 11.8%) or resolution (n = 1, 5.9%). At 3 years, the majority of cases had full resolution of initial neurologic symptoms (n = 9, 52.9%), and many had stable symptoms (n = 5, 29.4%). Detailed case descriptions are represented in table 2.

Table 2.

Exposed Cases With Subsequent Neuroinflammation

Discussion

Recent studies continue to caution use of TNFi over concerns for secondary demyelinating complications¹²; however, much of this caution is based on limited data from various case series. This study aimed to evaluate the risk of neuroinflammatory effects with these common medications by comprehensively studying a large cohort of TNFi-exposed subjects. Because TNFi-associated demyelination has been associated with heterogeneous neuroinflammatory conditions, we used a broad definition of neuroinflammation for this analysis rather than incorporating strict McDonald criteria for MS. This study suggests that neuroinflammatory complications are rare, occurring in less than 5% of those with TNFi exposure and a neurologic complaint precipitating a brain MRI. Moreover, only approximately 1 in 9 TNFi-treated patients in this cohort required brain imaging for a neurologic complaint. When calculating the percentage of neuroinflammatory disease in all patients at our institution exposed to TNF blockade over a 10-year period, the period prevalence was very low, at 0.4%.

Although IBD was the most common autoimmune disease diagnosis in our case group, we interpret this trend with caution, as another recent study found higher cases of rheumatoid arthritis in those with TNFi-associated neuroinflammation.¹³ Genetic susceptibility studies suggest that MS has significant overlap with IBD but also with many other autoimmune diseases.^14,15 It has been suggested that there is an increased risk of up to 4-fold of developing demyelinating disease with IBD, irrespective of TNF blockade. It has been speculated that the risk might also increase with TNF blockade; however, prior data did not reach statistical significance.¹⁶

Based on a cohort of 19 cases treated with infliximab or etanercept, Mohan et al. recommended avoiding TNFi in patients with a preexisting diagnosis of MS and using caution in those with a strong family history of MS. Because MS will inherently reveal demyelination on brain MRI, we did not include patients with an existing diagnosis of MS and TNFi exposure. The percentage of MRIs with neuroinflammation between the unexposed and exposed groups (3.8% and 4.5%) was similar and raises the question whether the MRI changes were truly from TNFi or rather demonstrating a baseline risk for neuroinflammation in people with systemic autoimmune disease. Of interest, in several of our cases, we observed that the form of neuroinflammation varied from that of classic MS. Specifically, it differed in terms of having an older age at onset, variable response to traditional MS treatments, and in some cases much more impressive degrees of refractory CNS inflammation. This could be due to a neurotoxic effect associated with TNF blockade, as opposed to traditional MS.

Additional research will be needed as the small number of cases limits the conclusions that can be drawn. Moreover, recent pathology studies reported that lesions from infliximab-associated CNS demyelinating disorders were indistinguishable from MS lesions.⁵ Although several theories have been proposed,^17–19 the mechanism of TNFi-associated demyelination is not understood, and further exploration is needed.

Limitations of the study include those inherent to a retrospective study, including dependence on accurate chart documentation. There was noted to be limited documentation of family history in many of the charts, and some patients were lost to follow-up in our medical record system. We were also unable to determine whether 1 type of TNFi was more strongly associated with neuroinflammation, given the low numbers of cases and frequent serial exposure to multiple TNFi among patients. In addition, a few cases had either remote exposure or short duration of exposure to TNFi. The significance of this exposure timing is unknown. Cumulative exposure to TNFi in months was not calculated due to data availability and use of multiple TNFi over the study period.

Furthermore, although all of the MRI studies were completed within a single system, the actual MRI machine and study protocols may have varied between study participants. Because of the retrospective nature of the study and the limited availability of documented testing, cerebrospinal fluid profiles were not systematically collected and included as part of the study parameters. They were, however, reported where available.

This study cohort was generated at a tertiary referral center, geographically located in the northeastern United States, which is an area with a known higher prevalence of autoimmune diseases. Moreover, because only MRI brain imaging was assessed, any patients with primarily spinal cord inflammation would not have been captured. These caveats may affect the generalizability of our findings. The retrospective design also might have missed MRIs performed outside of the YNHHS on subjects exposed to TNFi during the 10-year period. Despite these limitations, this study provides information from large numbers of patients exposed to TNFi, allowing for some basic understanding of prevalence and exposure risks, which was not previously available.

Despite initial concerns for neuroinflammatory phenomena following TNFi exposure, TNFi-associated neuroinflammation was actually a rare finding in this cohort, with a 10-year period prevalence of 0.4%. Within the subset of those presenting with neurologic complaints necessitating MRI brain imaging, the 10-year period prevalence increased to 4.5%. This study reveals trends that might suggest an increased risk of neuroinflammation in those with IBD or positive family history of IBD; however, larger studies are needed to more accurately assess this risk.

Although secondary neuroinflammation might be a rare event, we continue to recommend neurologic assessment and imaging in patients with TNFi exposure and focal neurologic symptoms. The results of this study would suggest that concern for de novo secondary neuroinflammation should not be a reason for withholding necessary treatment with TNFi when medically indicated and alternatives are not available.

TAKE-HOME POINTS

• → TNFi are used to treat many autoimmune conditions, but reports of neuroinflammatory effects have been associated with their use.

• → Suspected TNFi-associated neuroinflammation was a rare finding in this study cohort.

• → In cases with neuroinflammation in the setting of TNFi exposure, presentations included classic MS but with variable responses to traditional MS treatments. Nontraditional presentations with treatment-refractory CNS inflammation were also observed.

Appendix. Authors

Study Funding

No targeted funding reported.

Disclosure

A. W. Yu and M. Pecsok report no disclosures relevant to the manuscript. E. E. Longbrake has served as a consultant for Genentech, Genzyme, Alexion, Biogen, Celgene /Bristol-Myers Squibb, and EMD Serono. She is supported by research funding from Race to Erase MS and NINDS K23NS107624. S. F. Wesley reports no disclosures relevant to the manuscript. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.