MGB-NMOSD
(2026)Objective
Compare the real-world efficacy, tolerability, and safety of rituximab vs. FDA-approved NMOSD disease-modifying therapies (C5 inhibitors [eculizumab/ravulizumab], inebilizumab, satralizumab) and off-label treatments (mycophenolate mofetil, azathioprine) in patients with NMOSD.
Study Summary
• 5-year relapse-free probability: 100% for C5 inhibitors, inebilizumab, and satralizumab vs. 69.7% for rituximab, 51.1% for MMF, and 35.3% for azathioprine
• Annualized relapse rate: 0 for C5 inhibitors, inebilizumab, and satralizumab vs. 0.08 for rituximab, 0.19 for MMF, and 0.34 for azathioprine
• Composite endpoint (relapse, serious infection, or treatment-limiting AE): C5 inhibitors HR 0.22 (0.05-0.67) vs. rituximab; azathioprine HR 2.33 (1.08-4.86) and MMF HR 1.75 (1.02-2.95) were worse than rituximab
• Serious infection incidence rate ratio vs. rituximab: 0.16 (95% CI 0.05-0.42) for C5 inhibitors; 13% of rituximab periods had hypogammaglobulinemia and 28% had serious infections
Intervention
Maintenance NMOSD disease-modifying treatments: rituximab vs. C5 inhibitors (eculizumab or ravulizumab), inebilizumab, satralizumab, mycophenolate mofetil, and azathioprine
Inclusion Criteria
Patients meeting 2015 IPND criteria for NMOSD, evaluated in person at Mass General Brigham between Jan 1, 2000 and June 30, 2024
Study Design
Arms: Six treatment groups (single-arm exposures within one cohort): rituximab (n=192), C5 inhibitors (n=21), inebilizumab (n=11), satralizumab (n=19), MMF (n=65), azathioprine (n=23)
Patients per Arm: 176 patients total (331 treatment periods); rituximab 192, MMF 65, azathioprine 23, C5 inhibitors 21, satralizumab 19, inebilizumab 11
Outcome
• Primary - ARR was 0 for C5 inhibitors/inebilizumab/satralizumab, 0.08 for rituximab, 0.19 for MMF, 0.34 for azathioprine
• Composite (relapse + SIAE + TLAE) favored C5 inhibitors (HR 0.22) but disfavored azathioprine (HR 2.33) and MMF (HR 1.75) vs. rituximab
• Safety - C5 inhibitors had the lowest serious infection rate (IRR 0.16 vs. rituximab); rituximab had hypogammaglobulinemia in 13% and serious infections in 28% of treatment periods
Clinical Question
Among patients with NMOSD, how do the recently approved NMOSD-specific therapies (C5 inhibitors, inebilizumab, satralizumab) and off-label agents (MMF, azathioprine) compare with rituximab in terms of relapse-free survival, annualized relapse rate, and the combined risk of relapse, serious infectious adverse events, and treatment-limiting adverse events?
Bottom Line
In a real-world cohort of 176 patients with NMOSD followed for a median of 9 years, FDA-approved NMOSD-specific therapies (C5 inhibitors, inebilizumab, satralizumab) were substantially more effective and safer than rituximab, whereas MMF and azathioprine were less effective and carried higher composite risk - supporting first-line use of approved NMOSD DMTs (especially C5 inhibitors) over rituximab, MMF, or azathioprine.
Major Points
- Largest single-network real-world NMOSD comparative effectiveness cohort to date (Mass General Brigham, 2000-2024) with 176 patients (86% AQP4+, 83% female) and 331 treatment periods analyzed across 6 maintenance therapies.
- 5-year relapse-free probability was 100% for C5 inhibitors (eculizumab/ravulizumab), inebilizumab, and satralizumab, compared with 69.7% for rituximab, 51.1% for MMF, and 35.3% for azathioprine.
- Adjusted (Cox-Firth) relapse hazard ratios vs. rituximab: C5 inhibitors 0.12 (95% CI 0.07-0.24), inebilizumab 0.22 (0.12-0.65), satralizumab 0.19 (0.11-0.42).
- Annualized relapse rates: 0 for C5 inhibitors, inebilizumab, and satralizumab; 0.08 for rituximab; 0.19 for MMF; and 0.34 for azathioprine - results were similar after IPTW adjustment and when restricted to AQP4+ patients.
- C5 inhibitors had the lowest serious infection rate (incidence rate ratio 0.16 vs. rituximab, 95% CI 0.05-0.42, P=0.0002); 13% of rituximab treatment periods had hypogammaglobulinemia, 28% had serious infections, and 34% had common infections.
- Composite endpoint of relapse, SIAE, or TLAE favored C5 inhibitors (HR 0.22, 95% CI 0.05-0.67) and disfavored azathioprine (HR 2.33, 95% CI 1.08-4.86) and MMF (HR 1.75, 95% CI 1.02-2.95) compared with rituximab; inebilizumab and satralizumab were not significantly different from rituximab on the composite.
- Authors conclude clinicians should consider NMOSD-approved therapies first-line, avoid MMF and azathioprine, and caution against default first-line rituximab given cumulative relapse and adverse-event burden over time.
Study Design
- Study Type
- Retrospective real-world cohort study (comparative effectiveness)
- Randomization
- No
- Blinding
- Not blinded; dual neuroimmunologist chart review with senior-author adjudication of disagreements
- Sample Size
- 176
- Follow-up
- Median 9 years (IQR 5-14)
- Centers
- 1
- Countries
- USA
Primary Outcome
Definition: Relapse-free survival (time to first clinically definite or probable relapse) and annualized relapse rate (ARR), compared with rituximab
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| Rituximab: 5-year relapse-free probability 69.7%; ARR 0.08 (95% CI 0.062-0.10) | C5 inhibitors: 100% relapse-free at 5y, ARR 0 (95% CI 0-0.062); Inebilizumab: 100% relapse-free at 5y, ARR 0 (0-0.06); Satralizumab: 100% relapse-free at 5y, ARR 0 (0-0.17); MMF: 51.1% relapse-free at 5y, ARR 0.19 (0.14-0.26); Azathioprine: 35.3% relapse-free at 5y, ARR 0.34 (0.18-0.56) | 0.12 (C5 inhibitors), 0.22 (inebilizumab), 0.19 (satralizumab) vs. rituximab (C5: 0.07-0.24; Ineb: 0.12-0.65; Satra: 0.11-0.42) | All significant; ARR comparisons by Poisson and IPTW-adjusted negative binomial models |
Limitations & Criticisms
- Retrospective single-network (Mass General Brigham) design with non-randomized treatment assignment; despite IPTW adjustment for age and prior attacks, residual confounding is likely (e.g., disease severity, comorbidities, calendar-era of treatment).
- Tertiary referral ascertainment bias - cohort may overrepresent refractory or severe NMOSD, limiting generalizability.
- Marked differences in time on treatment by drug (rituximab median 3.4 y vs. satralizumab 0.6 y, inebilizumab 1.7 y) and small per-arm sample sizes (inebilizumab n=11, satralizumab n=19, C5 inhibitors n=21) limit statistical power and may underestimate late adverse events for newer drugs.
- Zero-event arms (C5 inhibitors, inebilizumab, satralizumab) required Firth penalized regression and exact Poisson methods; the composite endpoint can mask differential effects across efficacy vs. toxicity.
- Retrospective relapse assessment is subjective despite dual-reviewer adjudication; MRI confirmation availability varied (94% in definite, 9.5% confirmed in 'unlikely' events), and historical inconsistent monitoring limited longitudinal CD19/CD20 and IgG data.
- Inebilizumab and rituximab share B-cell depletion mechanisms - expected similar long-term safety profiles may emerge with longer follow-up; selection bias toward rituximab for patients ineligible for trials may also affect comparisons.
- Cost-effectiveness was not formally evaluated despite large price differences (C5 inhibitors >$500,000/year first year vs. rituximab biosimilars $10,000-$20,000/year); authors call for dedicated cost-effectiveness studies.
- Study funded by Alexion (AstraZeneca), manufacturer of the C5 inhibitors (eculizumab/ravulizumab) that were the top-performing arm - potential sponsor influence on framing/interpretation; several authors (including senior author S. Bhattacharyya and M. Levy) disclose Alexion-related funding or advisory relationships.
Citation
Neurol Neuroimmunol Neuroinflamm 2026;13(2):e200536