A Multi‐Center Retrospective Cohort Study of Neurosarcoidosis Myelitis: Current Observations and Future Directions

ABSTRACT

Objective

The optimal treatment for neurosarcoidosis myelitis is uncertain. We characterize incident neurosarcoidosis myelitis and assess treatment response by MRI and clinical scales.

Methods

Incident probable or definite neurosarcoidosis myelitis in adults was retrospectively identified from 13 academic medical centers. Cases were analyzed by initial treatment. The primary outcome was T1 post‐contrast gadolinium enhancement resolution at 6 months post‐treatment. Secondary outcomes were changes in modified Rankin scale (mRS) and Expanded Disability Status Scale (EDSS) from nadir to final follow‐up.

Results

Two hundred two patients were identified (median diagnosis age: 47 years (IQR 39–55); male: female 1.3:1). Median nadir mRS and EDSS were 2 (IQR 2–3) and 4 (IQR 2.5–6). At initial treatment, 129 (63.9%) received prolonged corticosteroids ≥ 4 weeks (group A₁), 36 (17.8%) received corticosteroids < 4 weeks (B₁), 21 (10.4%) received corticosteroids plus sarcoidosis‐directed immunosuppressant (E), and 16 (7.9%) received corticosteroids plus non‐sarcoidosis‐directed agents (F). In 167 cases with sufficient imaging, there were no significant differences in contrast enhancement resolution at 6 months (A₁ 27/106 (25.5%), B₁ 9/28 (32.1%), E 5/19 (26.3%), F 5/14 (35.7%); Fisher's exact p = 0.76). There were no significant differences in changes in mRS or EDSS among treatment groups (Kruskal–Wallis p = 0.69 and 0.63, respectively) after median follow‐up of 46.5 months (IQR 18–91.3).

Interpretation

Different initial immunosuppression strategies did not correlate with MRI contrast enhancement resolution at 6 months or clinical scales (mRS, EDSS). However, conclusions are limited by retrospective design, imbalanced cohorts, and insensitivity of binary MRI outcomes and available clinical scales for treatment response in neurosarcoidosis.

1. Introduction

Sarcoidosis is a systemic granulomatous inflammatory disease that commonly involves the lungs and lymph nodes, variably affecting other organ systems. Approximately 5%–35% of patients with sarcoidosis have neurologic involvement [1, 2]. In 50%–70% of neurosarcoidosis, neurologic symptoms are the first manifestation of sarcoidosis [1, 3]. Diagnosis may be elusive due to clinical heterogeneity and lack of reliable biomarkers. Tiers of diagnostic certainty are defined by collective interpretation of clinical presentation, imaging, serologic or cerebrospinal fluid (CSF) testing, exclusion of alternative causes, and in definite cases, biopsy confirmation of non‐caseating granulomas in neural tissue [4]. The publication of neurosarcoidosis consensus diagnostic criteria in 2018 [4] has since provided an improved framework for evaluation.

Sarcoidosis, inclusive of neurosarcoidosis, is a corticosteroid‐responsive disease; corticosteroids are the first‐line treatment for acute relapses [2]. Although the use of corticosteroids for neurosarcoidosis is widely accepted, there remains marked variability in dosing and duration of use. Additionally, whether, which, and when to use steroid‐sparing agents require further investigation. A recent expert consensus publication supports the early use of steroid‐sparing agents dependent upon neurosarcoidosis phenotype and severity [5]. Early use of steroid‐sparing agents may be beneficial for disease control while also lessening potential corticosteroid toxicities from prolonged use. This proposed framework of early combination immunotherapy challenges previous guidelines published by in the pulmonary literature, which advocate for a stepwise approach to treatment escalation [2, 6]. Prospective studies are lacking to reconcile these differing expert recommendations. Both groups agree that steroid‐sparing agents may not be necessary for all phenotypes, as some are sufficiently steroid‐responsive or often monophasic [5, 6]. Steroid‐sparing agents used in neurosarcoidosis include methotrexate, azathioprine, mycophenolate mofetil, cyclophosphamide, infliximab, and adalimumab [2, 3, 5, 6]. Infliximab has demonstrated efficacy against refractory and/or severe neurosarcoidosis [7, 8, 9]. Other immunotherapies have been reported for neurosarcoidosis, but these lack robust evidence of efficacy: rituximab, tofacitinib, and tocilizumab [10, 11, 12].

Rather than studying neurosarcoidosis in aggregate, analyzing treatment response by specific phenotype may yield the greatest insight. In this multicenter, retrospective study, we characterize neurosarcoidosis myelitis with attention to the assessment of response to varying treatment approaches. We hypothesize that the initial use of corticosteroids plus a sarcoidosis‐directed steroid‐sparing agent yields improved imaging and clinical responses as compared to corticosteroid monotherapy for incident, probable, and definite neurosarcoidosis myelitis.

2. Methods

A retrospective cohort of incident neurosarcoidosis myelitis from 13 institutions in the United States was collated for this study. Inclusion criteria required: age ≥ 18 years, clinical signs/symptoms of myelopathy, relevant spinal cord MRI abnormalities, incident neurosarcoidosis myelitis, no active receipt of sarcoidosis‐directed immunosuppression at onset (i.e., did not receive immunomodulating or immunosuppressive therapy inclusive but not limited to corticosteroids, methotrexate, mycophenolate mofetil, TNF‐alpha inhibition), meeting diagnostic criteria [4] for probable or definite neurosarcoidosis, and treatment with corticosteroids at a minimum. Excluded were possible neurosarcoidosis, concurrent other neurosarcoidosis clinical phenotypes, and treatment with immunosuppressants for comorbidities at myelitis onset. Sufficient pre‐ and post‐treatment imaging data, requiring use of gadolinium contrast, were needed for primary outcome analysis. Sufficient pre‐ and post‐clinical exam data to permit retrospective modified Rankin Scale (mRS) [13] or Expanded Disability Status Scale (EDSS) [14] scores were required for secondary outcome analysis.

Each institution performed retrospective chart reviews with extraction of defined clinical (demographics, comorbidities, signs/symptoms at incident event), imaging (lesion length, MRI contrast‐enhancement pattern; pulmonary imaging), serologic, cerebrospinal fluid, pathology, treatment, infectious complications, and response to treatment data. Each participating institution obtained approval for involvement in this collaborative study from respective institutional review board committees at each center. Of note, CSF cell count, glucose, and protein values were interpreted per the performing institution's assay thresholds. Retrospective mRS and EDSS scores were computed at nadir of incident myelitis and last post‐treatment follow‐up. To standardize retrospective mRS and EDSS scoring, each site received instructive documents (Supporting Information Materials) [13, 14]. Data from all institutions were de‐identified, then centrally collected and analyzed by the Mass General Brigham neurosarcoidosis research team.

Cases were grouped by initial treatment approach at the time of diagnosis: (A) intravenous corticosteroids followed by a prolonged oral corticosteroid taper (≥ 4 weeks), (B) intravenous corticosteroids followed by a short oral corticosteroid taper (< 4 weeks), (C) prolonged oral corticosteroid taper only (≥ 4 weeks), (D) short corticosteroid taper only (< 4 weeks), (E) corticosteroids PLUS a sarcoid‐directed steroid‐sparing immunosuppressive therapy (IST) (i.e., azathioprine, mycophenolate mofetil, methotrexate, cyclophosphamide, infliximab), and (F) corticosteroids PLUS other (others included: hydroxychloroquine, intravenous immunoglobulin (IVIg), glatiramer acetate). The minimum starting dose of corticosteroids considered as a therapeutic taper was oral prednisone 20 mg daily (note: most oral prednisone tapers began at 60–80 mg/day).

2.1. Statistical Analysis

For analysis, treatment groups (A–F) were re‐organized as follows: groups A and C were combined into the prolonged corticosteroid course (≥ 4 weeks) group, A ₁ . Groups B and D were combined into the short corticosteroid course (< 4 weeks) group, B ₁ . Groups E and F remained unchanged.

2.2. Primary Outcome

The primary outcome was assessed for complete resolution of gadolinium contrast enhancement on MRI of the pertinent spinal cord lesion within 6 months post‐treatment. This was analyzed as a binary outcome and compared across treatment groups A ₁ versus B ₁ versus E versus F using a Fisher's exact test. A defined 6‐month (≥ 6 months and < 7 months) timepoint was selected for analysis due to variable durations to post‐treatment imaging. Cases without pre‐treatment contrast enhancement on MRI were excluded from analysis.

2.3. Secondary Outcome

The secondary outcome assessed retrospective delta mRS and EDSS scores, respectively, among treatment groups (A ₁ , B ₁ , E, F) using a Kruskal–Wallis test. ΔmRS or ΔEDSS was the difference between nadir scores at initial presentation and final follow‐up.

2.4. Additional Sub‐Analyses

Kruskal–Wallis tests were performed to assess associations between treatment approach and disease severity (measured by nadir mRS and nadir EDSS, respectively), and separately with clinical remission occurrence. Clinical remission was defined as ≥ 3 months without clinical progression. Kruskal–Wallis tests were used to analyze associations between change in functional scores (ΔmRS, ΔEDSS) and duration to treatment start. Fisher's exact tests compared case characteristics (i.e., myelitis lesion length, MRI spine enhancement pattern, CSF pleocytosis presence) with the primary imaging outcome (contrast resolution on MRI by 6 months), and clinical/imaging relapse occurrences by treatment group. Clinical relapse was defined as persistent clinical worsening attributable to neurosarcoidosis myelitis following a prior ≥ 3‐month period of clinical remission. Imaging relapse was defined as the re‐appearance of contrast enhancement on post‐treatment MRIs after documented prior complete resolution.

3. Results

Two hundred two unique cases of incident definite or probable neurosarcoidosis myelitis in individuals ≥ 18 years of age, that were retrospectively identified from 13 institutions across the United States, met study inclusion criteria (Table S1). Of these, 167 had sufficient imaging data for primary outcome analysis. Those excluded had insufficient pre‐ or post‐treatment MRI data (n = 24), did not undergo pre‐treatment MRI studies with gadolinium contrast administration (n = 2), did not undergo an initial post‐treatment MRI study with gadolinium contrast administration (n = 1), or did not report/demonstrate contrast enhancement on the pre‐treatment MRI (n = 8; of these, 5 had short segment non‐enhancing lesions and 3 had longitudinally extensive myelitis lesions seen on FLAIR MRI sequences). One hundred seventy‐five were included in the secondary outcome analysis; 27 were excluded due to insufficient clinical data (Figure 1).

FIGURE 1.

Variable appearance of neurosarcoidosis myelitis on MRI. (A–D) Demonstration of neurosarcoidosis myelitis presenting as a longitudinal‐extensive transverse myelitis with variable enhancement patterns. (A) Sagittal STIR sequence displaying intramedullary hyperintensity spanning C4‐T3 with associated cord expansion. (B) Sagittal post‐contrast T1 sequence of the lesion displayed in A, which shows leptomeningeal and patchy intramedullary parenchymal enhancement of the spinal cord at levels C4, C5‐6, C7‐8. (C) Sagittal STIR sequence displaying a different longitudinally extensive transverse myelitis due to neurosarcoidosis with involvement of the corticomedullary junction through C7. (D) Sagittal T1 post‐contrast sequence showing patchy, intramedullary gadolinium‐contrast enhancement of the involved cervical spine lesion from panel C. (E, F) Demonstration of neurosarcoidosis myelitis presenting as a short segment lesion. (E) Sagittal STIR sequence displaying an intramedullary hyperintensity spanning levels T2–T3. (F) Axial post‐contrast T1 sequence displaying gadolinium contrast enhancement of the short segment myelitis lesion shown in panel E. (G) An axial T1 post‐contrast MRI sequence of a neurosarcoidosis myelitis lesion displaying the “Trident Sign” [15].

3.1. Demographics

The median age at neurosarcoidosis myelitis diagnosis for the entire cohort of 202 cases was 47 years (IQR 39–55), and the male to female ratio was 1.3:1 (Table 1). Myelitis was the first presentation of sarcoidosis in 182 (90.1%); the median age at systemic (non‐neurologic) sarcoidosis diagnosis was 45 years (IQR 40–53.35) for 86 cases with available data. Of those with known systemic sarcoidosis prior to neurologic involvement, neurosarcoidosis myelitis occurred within a median of 6 years (IQR 1–14) following the diagnosis of systemic, non‐neurologic sarcoidosis. 170 (84.2%) met criteria for “probable neurosarcoidosis” (positive biopsy of non‐neural tissue) and 32 (15.8%) met criteria for “definite neurosarcoidosis” (positive biopsy of neural tissue) [3]. Reported systemic biopsy sites included: lymph nodes (n = 146; including mediastinal, hilar, thoracic, paratracheal, supraclavicular, and mandibular), lung parenchyma/nodules (n = 12), skin (n = 5), ocular (n = 4; including conjunctiva, orbital mass, and lacrimal gland), liver (n = 2), and sinus (n = 1). Two of the definite neurosarcoidosis cases were isolated neurosarcoidosis without any systemic disease; biopsy confirmation was via procurement of spinal cord tissue.

TABLE 1.

Neurosarcoidosis myelitis collective cohort demographics and clinical data at incident presentation.

Median age at incident NS myelitis presentation, years	47 (IQR 39–55)
Sex
Females, n (%)	88 (43.6)
Males, n (%)	114 (56.4)
Race
Black, n (%)	90 (44.5)
White, n (%)	101 (50.0)
Asian, n (%)	0
Other, n (%)	11 (5.4)
Ethnicity
Hispanic, n (%)	1 (0.5)
Non‐hispanic, n (%)	189 (93.6)
Not reported, n (%)	12 (5.9)
NS diagnostic classification [3]
Probable, n (%)	170 (84.2)
Definite, n (%)	32 (15.8)
Median clinical functional nadir at incident event
mRS score	2 (IQR 2–3)
EDSS score	4 (IQR 2.5–6)
Elevated serum ACE of 138 tested, n (%)	23 (16.7)
CSF analysis
Pleocytosis (> 5 WBCs/μL), of 162 tested, n (%)	128 (79.0)
Elevated Protein, of 164 tested, n (%)	126 (76.8)
Elevated CSF ACE, of 51 tested, n (%)	10 (19.6)
Initial neuroimaging findings
Myelitis length
Longitudinally extensive, n (%)	134 (66.3)
Short segment, n (%)	66 (32.7)
Not reported, n (%)	2 (1.0)
Post‐T1 gadolinium contrast enhancement pattern
Parenchymal spinal cord enhancement only, n (%)	94 (46.5)
Leptomeningeal spinal cord enhancement only, n (%)	11 (5.4)
Both parenchymal and leptomeningeal enhancement, n (%)	69 (34.2)
Inadequate data to determine pattern of reported enhancement, n (%)	18 (8.9)
No contrast enhancement reported/demonstrated, n (%)	8 (4.0)
Insufficient Pre‐Treatment Imaging (i.e., no contrast administration), n (%)	2 (1.0)
Duration from NS myelitis diagnosis to initial treatment start, median (days)	90 (IQR 30–180)
Initial treatment regimen
A: IV corticosteroids then prolonged oral taper (≥ 4 weeks), n (%)	79 (39.1)
B: IV corticosteroids then short oral taper (< 4 weeks), n (%)	30 (14.9)
C: Prolonged oral corticosteroid taper only (≥ 4 weeks), n (%)	50 (24.8)
D: Short oral corticosteroid taper only (< 4 weeks), n (%)	6 (3.0)
E: Corticosteroids PLUS sarcoidosis‐directed steroid‐sparing IST a , n (%)	21 (10.4)
F: Corticosteroids PLUS other b , n (%)	16 (7.9)

^a Sarcoidosis‐directed IST included azathioprine, mycophenolate mofetil, methotrexate, cyclophosphamide, infliximab.

^b Other steroid‐sparing agents included hydroxychloroquine, intravenous immunoglobulin G (IVIg), glatiramer acetate.

3.2. Neurosarcoidosis Myelitis Features

At nadir of initial presentation, the associated disability for the entire cohort was mild to moderate (median nadir mRS 2[ IQR 2–3]; median nadir EDSS 4 [IQR 2.5–6]). Other symptoms were sensory abnormalities (n = 179, 88.6%), associated pain (n = 137, 67.8%), and bowel/bladder dysfunction (n = 107, 53.0%). Related erectile dysfunction was not recorded. Pre‐treatment imaging data pertaining to lesion length was available for 200 of the 202 cases (Tables 1 and 2). Longitudinally extensive transverse myelitis lesions traversing ≥ 3 vertebral segments (n = 134/200, 67.0%) were more common than short segment lesions (n = 66/200, 33.0%). Areas of spinal cord involvement were as follows: cervical and thoracic (n = 95), cervical only (n = 51), thoracic only (n = 33), cervical and thoracic and lumbar (n = 11), thoracic and lumbar (n = 8), cervical and lumbar (n = 2), and lumbar only (n = 2). Observed spinal cord enhancement patterns were parenchymal (n = 163), leptomeningeal (n = 80), or both (n = 69) (Table 1). Cauda equina or conus involvement was reported in 14 cases. Structural degenerative spinal disc disease was noted at the level of neurosarcoidosis myelitis in 61/200 (30.5%), although the severity of the degenerative changes was not recorded in this dataset.

TABLE 2.

Total cohort demographics and clinical data as organized by treatment group.

Variables	Total n = 202	Treatment group
		A 1	B 1	E	F
		n = 129 (63.9)	n = 36 (17.8)	n = 21 (10.4)	n = 16 (7.9)
Median Age at NS Myelitis Presentation (years)
n (IQR)	—	47.0 (39.0–56.0)	46.5 (39.8–54.3)	45 (41.0–53.0)	50.5 (38.5–56.3)
Sex, n (%)
Females	88 (43.6)	54 (41.9)	18 (50.0)	9 (42.9)	7 (43.8)
Males	114 (56.4)	75 (58.1)	18 (50.0)	12 (57.1)	9 (56.3)
Race, n (%)
Black	90 (44.6)	58 (45.0)	15 (41.7)	13 (61.9)	4 (25.0)
White	101 (50.0)	64 (49.6)	20 (55.6)	6 (28.6)	11 (68.8)
Asian	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
Other	11 (5.4)	7 (5.4)	1 (2.8)	2 (9.5)	1 (6.3)
Ethnicity, n (%)
Hispanic	1 (0.5)	1 (0.8)	0 (0.0)	0 (0.0)	0 (0.0)
Non‐hispanic	189 (93.6)	119 (92.2)	35 (97.2)	20 (95.2)	15 (93.8)
Not reported	12 (5.9)	9 (7.0)	1 (2.8)	1 (4.8)	1 (6.3)
NS diagnostic classification, n (%)
Probable	170 (84.2)	106 (82.2)	32 (88.9)	18 (85.7)	14 (87.5)
Definite	32 (15.8)	23 (17.8)	4 (11.1)	3 (14.3)	2 (12.5)
Median nadir clinical function
mRS (IQR)	—	2 (2–3)	2 (2–3)	3 (1–4)	3 (2–3.3)
EDSS (IQR)	—	4 (3–6)	3 (2–4.5)	4.8 (1.9–6.6)	4.3 (3–5.3)
Median duration from diagnosis to treatment start (months)
n (IQR)	—	3 (1–6)	3 (1.0–7.3)	1 (0.4–4.0)	1.8 (1.0–3.3)
Objective data of cases with available relevant data
Elevated serum ACE, n (%)	23/138 (16.7)	13/23 (56.5)	6/23 (26.1)	2/23 (8.7)	2/23 (8.7)
CSF analysis, n (%)
Pleocytosis	128/162 (79.0)	81/128 (63.3)	18/128 (14.1)	14/128 (10.9)	15/128 (11.7)
Elevated protein	126/164 (76.8)	77/126 (61.1)	20/126 (15.9)	18/126 (14.3)	11/126 (8.7)
Neuroimaging, n (%)
LETM	134/200 (67.0)	91/134 (67.9)	18/134 (13.4)	14/134 (10.4)	11/134 (8.2)
Short segment	66/200 (33.0)	36/66 (54.5)	18/66 (27.3)	7/66 (10.6)	5/66 (7.6)
Not reported	2	2	0	0	0

Abbreviation: LETM, longitudinally extensive transverse myelitis.

3.3. Supportive Objective Data

Serum angiotensin converting enzyme (ACE) was elevated in 23 (16.7%) of 138 tested cases. In the CSF, the median white blood cell (WBC) count was 18 cells/μL (IQR 8–49). Most displayed inflammatory CSF: 128 (79.0%) of 162 reported cases revealed pleocytosis, and 126 (76.8%) of 164 reported cases had elevated protein. CSF glucose was elevated in 43/145 (29.7%), normal range in 88/145 (60.7%), low in 14/145 (9.7%). CSF‐unique oligoclonal bands (OCBs) were elevated in 31/128 (24.2%), and CSF ACE was elevated in 10/51 (19.6%) of tested samples. Three cases had normal CSF cell count, normal CSF protein, but positive CSF oligoclonal bands.

3.4. Acute Treatment and Response

The median duration from symptom onset to treatment start, irrespective of approach, was 90 days (IQR 30–180) for the entire cohort. Prolonged corticosteroid course (A ₁ ) was received by 129 (63.9%) patients. Of these, 61.2% completed a preceding 3–5‐day intravenous methylprednisolone course. Short corticosteroid course (B ₁ ) was used to treat 36 (17.8%) patients, and of these, 30 (83.3%) completed a preceding high‐dose intravenous methylprednisolone course. Corticosteroids plus a sarcoidosis‐directed steroid‐sparing IST (E) were administered as initial, acute treatment for 21 (10.4%) patients. The sarcoidosis‐directed steroid‐sparing agents received by the 21 total cases of treatment group E were as follows: intravenous infliximab 3–5 mg/kg given at least once with several reporting a frequency of every 4–6 weeks (n = 8), oral methotrexate weekly with a dose range of 7.5‐25 mg or dosing not specified (n = 7), oral mycophenolate mofetil 1000 mg twice daily (n = 3), oral azathioprine 200–300 mg daily (n = 2), intravenous cyclophosphamide transitioned to oral mycophenolate mofetil twice daily without recorded dosing (n = 1). In a minority (n = 16, 7.9%), other treatments, not thought to be sarcoidosis‐directed (F), were administered along with corticosteroids: therapeutic plasma exchange 5–6 sessions (n = 8), IVIg dosing not reported (n = 2), hydroxychloroquine dosing not reported (n = 2), glatiramer acetate (n = 2), IVIg dosing not reported and therapeutic plasma exchange 5 sessions (n = 1), and specific non‐sarcoidosis other not reported (n = 1). There was no association found between nadir clinical function scores at initial presentation and treatment approach for the entire cohort (nadir EDSS Kruskal–Wallis p = 0.11; nadir mRS Kruskal–Wallis p = 0.13).

The imaging and clinical responses to treatment are reported for patients with sufficient data to permit analysis (n = 167 had sufficient imaging data; n = 175 had sufficient clinical data). Of the 167 cases with sufficient data for primary outcome analysis, the median duration from treatment start to first post‐treatment MRI was 90 days (IQR 46.5–154.5). There were no significant differences in the proportions of patients with contrast enhancement resolution at 6 months post‐treatment among individual treatment strategies or treatment groups (Table 3; Fisher's exact p = 0.69 and p = 0.76, respectively). Those with partial contrast resolution were tallied as lack of resolution. Complete resolution of gadolinium‐contrast enhancement on post‐treatment imaging (imaging remission) was reported in 136/152 (89.5%) cases, [A ₁ 86/94 (91.5%), B ₁ 22/25 (88.0%), E 15/19 (78.9%), and F 13/14 (92.9%); Fisher's exact p = 0.39]. Note, any changes in immunosuppression regimen after the initial treatment course were not recorded. Following imaging remission, continued imaging surveillance occurred in 120 cases, of whom 37 (30.8%) reported return of contrast enhancement (imaging relapse). The relation to any change in immunosuppressive regimen during this period of surveillance was not recorded.

TABLE 3.

Neurosarcoidosis myelitis imaging outcomes by treatment approach for 167 cases with sufficient data.

Treatment group	Median duration from symptom onset to treatment start, months	MRI contrast resolution by 6 months, yes n, (%)	Median duration from treatment start to MRI contrast resolution, months
A 1 Prolonged corticosteroid taper (n = 106)	3.0 (IQR 1.0–6.0)	27/106 (25.5)	9.9 (IQR 5.0–17.7)
B 1 Short corticosteroids taper (n = 30)	2.5 (IQR 1–5.5)	9/28 (32.1)	7.0 (IQR 3.0–20.0)
E Corticosteroids PLUS sarcoidosis‐directed IST (n = 19)	1.0 (IQR 0.5–3.5)	5/19 (26.3)	14.0 (IQR 6.0–20.8)
F Corticosteroids PLUS other (n = 14)	1.8 (IQR 1.0–3.8)	5/14 (35.8)	9.0 (IQR 4.0–39.0)
Total (all groups), n = 167 with sufficient imaging data	2.4 (IQR 1.0–6.0)	46/167 (27.5)	10.0 (IQR 4.4–20.0)

Clinical remission occurred in 155/175 (88.6%) of cases with sufficient post‐treatment clinical data, with a median duration from treatment to clinical remission of 4 months (IQR 3–6) and a median duration of follow‐up from initial treatment to last clinical evaluation of 47.1 months (IQR 18.2–95.1). There were no significant differences among treatment groups for the duration from initial treatment to clinical remission (Kruskal–Wallis, p = 0.96), Table 4. Of the 155 cases with clinical remission, 60 reported clinical relapses thereafter, without association with the initial treatment group [A ₁ 37/100 (37.0%), B ₁ 10/25 (40.0%), E 8/16 (50.0%), F 5/14 (35.7%), Fisher's exact, p = 0.17]. There were no differences in final or change in clinical function scores among treatment groups (ΔEDSS Kruskal–Wallis p = 0.63; ΔmRS Kruskal–Wallis p = 0.69; final EDSS Kruskal–Wallis p = 0.40; final mRS Kruskal–Wallis p = 0.33), Table 4. Ambulatory status at final follow‐up was reported for 88 cases, 45/88 (51.1%) ambulated without assistance. There was no association between final ambulatory status and treatment group (Fisher's exact, p = 0.36), Table 4.

TABLE 4.

Neurosarcoidosis myelitis clinical outcomes by treatment approach for 175 cases with sufficient data.

Treatment group	Median duration from treatment start to last evaluation, months	Median nadir clinical function, mRS EDSS	Median change in clinical function Delta mRS Delta EDSS	Clinical remission reported, n (%)	Ambulatory at last evaluation, n (%)
A 1 Prolonged corticosteroid taper, n = 109	58 (IQR 18.8–96)	mRS 2 (IQR 1–3) EDSS 3.5 (IQR 2.5–6.0)	Delta mRS 0 (IQR −1‐1) Delta EDSS 1 (IQR 0–3)	100 (91.7)	27 (24.8)
B 1 Short corticosteroids taper, n = 29	34.8 (IQR 16–72.3)	mRS 2 (IQR 2–2) EDSS 3 (IQR 2–4.5)	Delta mRS 0 (IQR −1‐0) Delta EDSS 2 (IQR 0–4)	25 (86.2)	7 (24.1)
E Corticosteroids PLUS sarcoidosis‐directed IST, n = 21	46 (IQR 18.7–93.8)	mRS 3 (IQR 1–4) EDSS 4 (IQR 2–7)	Delta mRS 0 (IQR 0–1) Delta EDSS 1 (IQR 0–2)	16 (76.2)	6 (28.6)
F Corticosteroids PLUS other, n = 16	61.3 (IQR 17.1–86.6)	mRS 2 (IQR 1–3) EDSS 3.5 (IQR 2.9–5.1)	Delta mRS 0 (IQR −2‐0.3) Delta EDSS 1 (IQR 0.4–5.3)	14 (87.5)	5 (31.3)
Total (all groups), n = 175 with sufficient clinical data	47.1 (IQR 18.2–95.1)	mRS 2 (IQR 1–3) EDSS 3.5 (IQR 2.5–6)	Delta mRS 0 (IQR −1‐1) Delta EDSS 1 (IQR 0–3)	155 (88.6)	45 (25.7)

Outcomes were further analyzed by duration to treatment start within or greater than 6 weeks from symptom onset, and no significant differences in imaging or clinical outcomes were found (Supporting Information Tables). Additional analyses comparing CSF characteristics with outcomes are shown in Supporting Information Tables. The duration between symptom onset or treatment start and lumbar puncture was not recorded.

3.5. Treatment‐Related Complications

Thirty‐four (16.8%) of the total 202 cases reported infectious complications, including varicella zoster, persistent oral candidiasis, recurrent urinary tract infections (UTIs), sacral ulcer infection, bronchitis, and pneumonia, amongst others. None reported progressive multifocal leukoencephalopathy. Rates of infectious complications did not vary by treatment group (Fisher's exact, p = 0.99). Corticosteroid‐related toxicities were not explicitly recorded.

4. Discussion

Neurosarcoidosis, an inflammatory condition with clinical variability, lacks a standardized treatment approach. Published diagnostic criteria in 2018 [4] provided a framework to enable collaborative, retrospective efforts to characterize neurosarcoidosis phenotypes. Such retrospective studies may provide reliable characterizations; however, conclusions regarding optimal treatment approach must be cautiously interpreted. No prospective, randomized controlled trials have been performed to date in neurosarcoidosis, and clinical practice remains variable.

In this collaborative, multi‐institutional study, we comprehensively characterized incident neurosarcoidosis myelitis and analyzed our approach to treatment through retrospective assessment of imaging and clinical outcomes. Many descriptive characteristics of this cohort align with those published. We found a slight male predominance in this cohort (male: female 1.3:1) and a median age of onset of neurosarcoidosis myelitis of 47 years, similar to other published cohorts [16, 17, 18]. Interestingly, this ~1:1 or slight male‐predominant sex ratio interestingly conflicts with the observed female predominant (2:1) sex ratio of systemic sarcoidosis in the United States [19]. Akin to other published neurosarcoidosis myelitis cohorts, longitudinally extensive transverse myelitis and CSF pleocytosis were common [17, 18], although short segment lesions and normal cell counts were reported.

While most with this condition display inflammatory CSF via pleocytosis and/or elevated CSF protein, it is important to test for other markers of inflammation as well, specifically for CSF OCBs. In our cohort, CSF OCBs were elevated in ~24%, similar to other reports [18], with three of these cases having elevated OCBs with normal CSF cell count and protein. It is important to note this possibility of elevated CSF OCBs in neurosarcoidosis myelitis to avoid misdiagnosis of multiple sclerosis. CSF ACE was found to be a poor predictor of neurosarcoidosis myelitis. These data highlight a lack of available, reliable biomarkers for neurosarcoidosis. Pathologic evidence of sarcoidosis, either non‐neurologic or neurologic, should be sought; lymph node biopsy is the most commonly obtained source. In our study, no correlations between myelitis length, gadolinium contrast enhancement pattern, CSF pleocytosis, or CSF glucose were found (Table S4). CSF glucose values and pleocytosis were also not found to be associated with nadir or final clinical functional outcomes in this cohort (Table S5).

In this cohort, the median duration from symptom onset to treatment was 90 days (IQR 30–180). This possibly reflects the diagnostic challenges inherent to neurosarcoidosis or may be representative of an indolent, progressive course. Our data did not capture the pace of symptom development. To assess whether time to treatment correlates with defined imaging and clinical outcomes, supplemental sub‐analyses stratified by time to treatment (≤ 6 weeks or > 6 weeks) were completed, and no differences by approach were found (Tables S2 and S3). Due to the limitations of our study, we cannot definitively draw conclusions pertinent to clinical and imaging outcome measures in response to initial treatment regimens administered in neurosarcoidosis myelitis. Rather, the limitations of our study design and available data largely permit the reporting of observations. This highlights an important need for a prospective study which would first require standardization of both treatment and assessments of response in neurosarcoidosis myelitis.

Corticosteroid monotherapy was most commonly used for acute, initial management. However, there was significant variability in dosing and duration of use. Available expert guidance agrees upon universal use of corticosteroids in neurosarcoidosis, but there remains debate as to whether corticosteroid monotherapy or corticosteroids with a steroid‐sparing immunosuppressant therapy (IST) are most appropriate as initial treatment [5, 6]. The European Respiratory Society recommends an escalation approach to treatment: corticosteroid monotherapy to start, with escalation to IST if lack of response or significant corticosteroid side effects prohibiting continued monotherapy [6]. In contrast, recent neuroimmunologist‐driven consensus guidelines advocated for early use of corticosteroids and a steroid‐sparing IST for moderate to severe neurosarcoidosis myelitis [5]. This warrants further investigation as no prospective trials have yet been performed in neurosarcoidosis.

Our study has limitations both in study design and from the nature of sarcoidosis itself. The clinical study endpoints of EDSS and mRS were retrospectively ascertained. Retrospective EDSS scores correlate well with prospectively collected data in multiple sclerosis [20], but similar studies in neurosarcoidosis do not exist. Our primary endpoint evaluated the resolution of gadolinium enhancement at 6 months. Resultant from the multicenter design of the study with separate imaging platforms, imaging could not be centrally evaluated to permit adjudication of the primary endpoint. Clinical interpretation of imaging was unblinded. The study would be improved by having a central rater of imaging blinded to treatment assignment. We evaluated the binary primary outcome as the presence or absence of gadolinium enhancement. This outcome assumes that gadolinium enhancement represents active inflammation, but injury to the blood‐spinal cord barrier or gliosis may also cause chronic gadolinium enhancement. Improvement of enhancement to varying degrees (such as > 50% improvement in enhancement) may more accurately represent treatment response rather than complete resolution. Further research correlating imaging to pathological features or markers of inflammation is needed to establish the threshold of imaging improvement that represents treatment response.

Our study was limited by uneven sample sizes in each treatment assignment. Particularly for the combination therapy group of corticosteroids and sarcoid‐directed treatments, we had relatively few patients. We grouped together treatments of varying efficacy (such as infliximab and mycophenolate mofetil), and our conclusions are limited regarding individual treatments. Clinical follow‐up of patients was not standardized among centers, and this variability directly affected when follow‐up imaging and clinical outcomes were measured. Finally, there is no disease severity scale that is validated to be sensitive to clinical changes in neurosarcoidosis. The outcome measures of EDSS and mRS are relatively insensitive, and other clinical series on neurosarcoidosis evaluating treatment similarly find modest changes in disease severity scales with follow‐up [16]. We also did not evaluate systemic sarcoidosis disease activity in our clinical outcome measures, which is clinically important to patient outcomes.

The limitations of our study stemmed heavily from the heterogenous state of care provided to individuals with neurosarcoidosis. This warrants a call for standardization of management. Prior to an ability to proceed with a much‐needed prospective, randomized controlled trial, there remains a need to better define appropriate imaging and clinical endpoints in neurosarcoidosis, likely requiring incorporation of systemic disease activity. As mentioned, there exists a need to better understand how gadolinium enhancement correlates with pathophysiologic disease activity prior to relying upon this modality as an imaging endpoint for treatment responsiveness. The acknowledgment of the insensitivities of EDSS and mRS for neurosarcoidosis, as further highlighted by discrepancies with clinical progression [16], prompts consideration of a sarcoidosis‐specific clinical outcome measure. We postulate that an effective clinical outcome measure in neurosarcoidosis may require an integration of patient‐reported outcome measures that assess quality of life with defined objective outcome measures of function (e.g., the Timed Up and Go, 9‐Hole Peg Test, Standing Balance Test) [21]. Considering current utilization of corticosteroids for neurosarcoidosis, a clinical measure would need to appropriately control for effects of prolonged corticosteroid exposure. The Glucocorticoid Toxicity Index, a prospective tool to measure steroid toxicity, would be useful to employ [22].

Contrary to neurosarcoidosis, pulmonary, cardiac, and dermatologic sarcoidosis have successfully performed clinical trials due to comparative increased prevalence, using steroid reduction as an endpoint and the existence of composite outcome measures [23, 24, 25]. While the neurosarcoidosis community could similarly devise its own composite measures, its rarity challenges trial recruitment and studying neurosarcoidosis alone may not best equate to meaningful outcomes to a patient who is likely to have concurrent, other organ manifestations of sarcoidosis. Thus, the assessment of multi‐organ composite endpoints, as is performed for other multisystem rheumatologic diseases such as ANCA‐associated vasculitis or IgG4‐Related Disease, may yield success for clinical trials in sarcoidosis irrespective of the degree of neurologic or non‐neurologic manifestations. In the interim, a collaborative pursuit of a national, multi‐disciplinary patient registry of sarcoidosis, structured with pre‐specified follow‐up timepoints for prospective collection of clinical, imaging, serum, CSF, bronchoalveolar lavage, and tissue data could sufficiently provide an informative framework to facilitate the discovery of multi‐system composite endpoints and biomarkers in sarcoidosis, with application for neurosarcoidosis.

5. Conclusions

Retrospective cohort studies of neurosarcoidosis phenotypes permit reliable clinical characterization, though variable clinical practice limits assessment of the optimal treatment approach. Limitations due to heterogeneity in practice and thus small subsets per treatment approach are thought to have contributed to the lack of significant differences found in imaging and clinical function outcomes when comparing by treatment approach, specifically corticosteroid monotherapy versus combination immunotherapy. This discrepancy, which conflicts with anecdotal expert experience and other published literature, highlights the important need for a prospective, randomized controlled trial to better discern the optimal treatment approach for neurosarcoidosis myelitis.

Author Contributions

G.S.M. played a critical role in overall project design, multi‐institutional data collection and analysis, affiliated institution data collection, manuscript drafting/editing/finalization, table and figure design. D.B. contributed substantially to project design, multi‐institutional data collection, affiliated institution data collection, manuscript editing and finalization. Y.Z. contributed to data collection, statistical analysis, manuscript editing and table design. B.K.C., M.L., N.V., B.J.S., C.A.P. contributed to project design, manuscript editing and finalization. P.B., N.B.R., E.P.F., V.R., A.J.A., S.H., M.H., S.E.S., E.C.R., L.S., H.R., M.M., T.R., S.M.R., L.H., S.C., J.L., T.A.C., L.A.W., J.R.B., R.D.S., N.R.T., D.B.C., S.R.D., M.M., A.Z., S.M., Y.M.‐D., J.D., P.B.A. each contributed to institutional data collection and manuscript editing. Images for the manuscript's figure were provided and edited by G.S.M., E.P.F., T.A.C., and J.R.B., and S.B. S.B. played a critical role in overall project design, multi‐institutional data collection and analysis, affiliated institution data collection, manuscript drafting/editing/finalization, and table and figure design.

Conflicts of Interest

B.H. has received research support from Novartis and Genzyme; these disclosures are unrelated to this study. M.L. receives personal compensation for advising the following companies: Alexion, Horizon, Genentech/Roche, UCB, Sanofi, and Mitsubishi. Through Mass General, Dr. Levy also received grants from Alexion, Horizon, Genentech, Roche, UCB, and Sanofi for research projects; no funding is pertinent to this project. Although this specific manuscript study was not funded by NIH, C.A.P. and B.J.S. have received funding from NIH for neurosarcoidosis research on phenotypes. B.J.S. has also received royalties from publications related to neurosarcoidosis. P.B. received funding from the Foundation for Sarcoidosis Research, and the American Academy of Neurology paid to the institution (JHH) during the time the data was generated. E.P.F. has served on advisory boards for Alexion, Genentech, Horizon Therapeutics, and UCB. He has received research support from UCB. He received royalties from UpToDate. Dr. Flanagan is a site principal investigator in a randomized clinical trial of Rozanolixizumab for relapsing myelin oligodendrocyte glycoprotein antibody‐associated disease run by UCB and is a site principal investigator and a member of the steering committee for a clinical trial of satralizumab for relapsing myelin oligodendrocyte glycoprotein antibody‐associated disease run by Roche/Genentech. Dr. Flanagan has received funding from the NIH (R01NS113828), is a member of the medical advisory board of the MOG project, and is an editorial board member for the following journals: Neurology, Neuroimmunology and Neuroinflammation, The Journal of the Neurological Sciences, and Neuroimmunology Reports. E.C.R. has served on advisory boards for Sanofi and Amgen. L.S. has served on an advisory board for Amgen. M.M. has served on advisory boards for Horizon and Amgen therapeutics. T.R. has received grant funding from the National Multiple Sclerosis Society. Dr. Rempe has served on advisory boards for Genentech, Alexion, Amgen, TG Therapeutics, EMD Serono, and Sanofi/Genzyme. He receives contract research support from Sanofi/Genzyme, Novartis, Celgene, EMD Serono, and Genentech. S.C. is involved in the following industry‐sponsored clinical trials: Alexion clinical trial for Eculizumab in Relapsing NMO patients (site investigator) and the UCB clinical trial for relapsing MOGAD patients (site investigator). Dr. Clardy receives research support from NIH/NINDS (U01, The ExTINGUISH Trial), the Siegel Rare Neuroimmune Association, the Immune Deficiency Foundation, Horizon/Amgen, Alexion/AstraZeneca, the Barbara Gural Steinmetz Foundation, and Sumaira Foundation for NMO. Dr. Clary serves as a consultant/advisory board member for Alexion/AstraZeneca, Horizon/Amgen, Arialys; additionally, as a Medical Advisory Board member of the Sumaira Foundation for NMO. Dr. Clardy is the Section Editor for the Neurology Podcast and Neurology Minute, as well as an Editorial Board member for Neurology: Neuroimmunology and Neuroinflammation. T.A.C. has received personal compensation for advising the following companies: Kyverna, Horizon, Delve Bio through University of Iowa Health Care; has also received funding from Roche and Sanofi for involvement in clinical trials. J.R.B. has received personal compensation for serving as a Consultant for Celgene/BMS, Cycle Pharma, Dice Therapeutics, Genentech/Roche, Gilead, Janssen/J&J, Merck, Morphic, Novartis, Sandoz, Seagen, Takeda, and TG Therapeutics; personal compensation for serving on a Scientific Advisory or Data Safety Monitoring board for MAPI and ExcisionBio; the institution of Dr. Berger has received research support from Genentech/Roche. D.B.C. has received consulting fees from Roche, Excision Biotherapeutics, Cellevolve, Seagen; has served on DSMB or Adjudication Committees for Teva, Wave Life Sciences, Sanofi/Genzyme, Atara, and Cellevolve Bio; receives research support from NIA, NINDS, NIMH, and NIAID. A.Z. has served on advisory boards and speakers bureau for the following: Acorda, Alexion, AstraZeneca, Biogen, Cellgen, BMS, Genentech, Sanofi/Genzyme, Novartis, Janssen, EMD Serono, and TG. Y.M.‐D. has served as a consultant for and/or received grant support from Acorda, Bayer Pharmaceutical, Biogen Idec, EMD Serono, Sanofi/Genzyme, Novartis, Questor, Roche‐Genentech, Bristol‐Myers Squibb, Horizon/Amgen, Teva Neuroscience, and NIH. S.B. has received research support from NIH, Alexion Pharmaceuticals, UCB, Roche, GlaxoSmithKline, Massachusetts Consortium on Pathogens; consulting fees from Teladoc Health, Merck, Alexion Pharmaceuticals, NeuroLambda; honoraria from UpToDate, Wiley, and the American Academy of Neurology.

Supporting information

Appendix A. Supplementary data and tables.

Data Availability Statement

Anonymized data not published within this article will be made available by request from any qualified investigator.