← Back
NeuroTrials.ai
Neurology Clinical Trial Database

MG-BNMA

Comparison of the Efficacy and Safety of Myasthenia Gravis Treatments: A Bayesian Network Meta-Analysis

Year of Publication: 2026

Authors: McLaren NP, Rosati M, Zhong W, ..., Roy B

Journal: Neurology

Citation: Neurology. 2026 Aug 11;107(3):e218351.

Link: https://doi.org/10.1212/WNL.0000000000218351

Bottom Line

FcRn inhibitors, C5 complement inhibitors, and CD19+ B-cell depletion therapy show comparable efficacy on QMG and MG-ADL reduction in gMG (differences between classes below minimum clinically important thresholds), and FcRn inhibitors carry higher odds of treatment-related adverse events than placebo, whereas C5i and CD19+ BCDT do not.

Major Points

  • Bayesian NMA of 27 placebo-controlled RCTs (1987–2025; 17 treatments across 9 mechanism classes) in generalized MG, including recently approved targeted biologics.
  • FcRn inhibitors produced the largest QMG reduction vs placebo (−3.63; 95% CrI −4.43 to −2.77), followed by C5 complement inhibitors (−2.59; −3.92 to −1.28) and CD19+ B-cell depletion (−2.50; −5.11 to 0.12).
  • MG-ADL reductions were similar across FcRn (−1.93), CD19+ BCDT (−1.91), and C5i (−1.84) — none exceeded the 2-point clinically meaningful difference between active arms.
  • Higher percentage of female patients and greater enrollment age were associated with greater QMG reduction (sex β = −3.16; age β = 2.09), suggesting population differences may confound between-trial comparisons.
  • In an AChR-antibody–positive subgroup analysis, FcRn inhibitors showed the greatest treatment effect (QMG −3.81; MG-ADL −2.20) with CD19+ BCDT and C5i showing broadly comparable magnitudes.
  • Treatment-related adverse events: FcRn inhibitors (OR 2.20; 1.52–3.38), IL-6 signaling inhibitors (OR 3.24; 1.29–8.29), and IVIg (OR 3.62; 1.01–13.21) had higher TRAE odds than placebo; CD20+ B-cell depletion (SUCRA 0.735) and IST (SUCRA 0.690) had the lowest odds of TRAE.

Design

Study Type: Bayesian Network Meta-Analysis of Randomized Controlled Trials

Randomization:

Blinding: N/A (meta-analysis; included trials were placebo-controlled and mostly double-blind)

Enrollment Period: Included trials published 1987–2025; database search through May 2, 2025

Follow-up Duration: Included trials 4–52 weeks (outcome time points at which QMG/MG-ADL treatment effects were reported)

Centers: Multiple (across 27 RCTs)

Countries: Multinational (varied by included trial; several US, European, and Chinese trials)

Sample Size: 2318

Analysis: Bayesian random-effects hierarchical model via gemtc R package; Markov Chain Monte Carlo; sensitivity meta-regression adjusting for age, sex, and baseline QMG/MG-ADL; SUCRA ranking; Cochrane RoB 2 risk-of-bias assessment; PROSPERO CRD42023428733; PRISMA-NMA reporting.


Inclusion Criteria

  • Prospective placebo-controlled randomized trials
  • Adults (18 years or older) with generalized myasthenia gravis
  • Reported mean treatment difference (with SE, SD, or CI) in MG-ADL or Quantitative MG (QMG) score change from baseline
  • Multiarm study design
  • Any of the following mechanism classes: CD20+ B-cell depletion, CD19+ B-cell depletion, C5 complement inhibitor, IVIg, FcRn inhibitor, CD40 inhibitor, IL-6 signaling inhibitor, BAFF inhibitor, or immunosuppressive therapy/steroid-sparing agent

Exclusion Criteria

  • Trials reporting median (rather than mean) change from baseline (excluded azathioprine and thymectomy pivotal trials)
  • Trials that did not report change from baseline in QMG or MG-ADL (e.g., 2008 international phase 3 mycophenolate mofetil trial)
  • IVIg trials with outcomes that could not be aligned across studies
  • Retrospective, single-arm, or non-placebo-controlled designs
  • Studies of pediatric MG or purely ocular MG

Baseline Characteristics

CharacteristicControl (Placebo + Standard of Care)Active (Any Investigational or Comparator Treatment)
Number of patients (pooled)1,0861,232
Mean age at enrollment (across studies)~50 years (study means 40–66)~50 years (study means 36–67)
Female (across studies)~57.5% (weighted average)~57.5% (weighted average)
Baseline MG-ADL score (mean across studies)7.87.8
Baseline QMG score (mean across studies)15.615.6
Number of RCTs contributing placebo arms27
MG subtypePredominantly AChR-antibody positive; several trials mixed AChR+/MuSK+/LRP4+/seronegative
Concomitant therapyStandard of care (varied: pyridostigmine, corticosteroids, non-steroidal ISTs depending on trial protocol)
Number of treatments evaluated17 (across 9 mechanism-of-action classes)
Study length (weeks)4–52 (median ~24)
Trial phase distributionPhase 2 and phase 3 RCTs (phase 3 subset used for confirmatory analyses)

Arms

FieldFcRn inhibitorsC5 complement inhibitorsCD19+ B-cell depletionCD20+ B-cell depletionIL-6 signaling inhibitorIVIgCD40 inhibitorBAFF inhibitorImmunosuppressive therapy / steroid-sparingControl
InterventionEfgartigimod, nipocalimab, rozanolixizumab, batoclimab (dosing varied by trial)Eculizumab, ravulizumab, zilucoplanInebilizumabRituximab (BeatMG, RINOMAX)Satralizumab (LUMINESCE)IV immunoglobulin maintenance therapyIscalimabBelimumabMycophenolate mofetil, tacrolimus, cyclosporine, methotrexateMatching placebo added to standard-of-care background therapy
Duration4–24 weeks per trial12–26 weeks26 weeks (with 52-week extension analysis)24–52 weeks24 weeks24 weeks25 weeks36 weeks12–52 weeks4–52 weeks (matches active arm in each RCT)

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Mean treatment difference in QMG score change from baseline vs placebo (all-phase, unadjusted)PrimaryPlacebo + standard of care (reference)FcRn inhibitors: −3.63 pts | C5i: −2.59 pts | CD19+ BCDT: −2.50 ptsBayesian NMA — significance interpreted by credible intervals excluding zero (FcRn and C5i significant; CD19+ BCDT crossed zero)
MG-ADL mean difference vs placebo (unadjusted)SecondaryPlaceboFcRn −1.93 ; CD19+ BCDT −1.91 ; C5i −1.84N/A (mean difference)
QMG mean difference — phase 3 subsetSecondaryPlaceboFcRn −4.03 ; CD19+ BCDT −2.5 ; C5i −2.42N/A
MG-ADL mean difference — phase 3 subsetSecondaryPlaceboFcRn −2.27 ; CD19+ BCDT −1.9 ; C5i −1.71N/A
QMG in AChR-antibody positive subgroupSecondaryPlaceboFcRn −3.81 ; C5i −2.50 ; CD19+ BCDT −2.50N/A
MG-ADL in AChR+ subgroupSecondaryPlaceboFcRn −2.20 ; CD19+ BCDT −1.79 ; C5i −1.77N/A
MG-ADL in MuSK-antibody positive subgroup (best treatment)SecondaryPlaceboRozanolixizumab 7 mg/kg −9.42N/A
CD19+ BCDT extended 52-week analysis (AChR+)SecondaryPlaceboQMG −4.3 ; MG-ADL −2.79N/A
SUCRA probability of best QMG/MG-ADL efficacy (FcRn inhibitors)SecondaryN/A~0.90N/ARanked highest across nearly all endpoints
Covariate adjustment: % female effect on QMG treatment effectSecondaryN/Aβ = −3.16N/A
Covariate adjustment: enrollment age effect on QMG treatment effectSecondaryN/Aβ = 2.09N/A
FcRn inhibitors — TRAE OR vs placeboAdverse2.20 (95% CrI 1.52–3.38)
IL-6 signaling inhibitor — TRAE OR vs placeboAdverse3.24 (95% CrI 1.29–8.29)
IVIg — TRAE OR vs placeboAdverse3.62 (95% CrI 1.01–13.21) — driven by a single trial, interpret cautiously
C5 complement inhibitors — TRAE OR vs placeboAdverseComparable to placebo (95% CrI includes 1)
CD19+ B-cell depletion — TRAE OR vs placeboAdverseComparable to placebo (95% CrI includes 1)
CD20+ B-cell depletion — TRAE OR vs placeboAdverse0.71 (95% CrI 0.15–3.23) — lowest odds (SUCRA 0.735)
BAFF inhibitor (belimumab) — TRAE OR vs placeboAdverse0.81 (95% CrI 0.17–3.84)
IST (MMF, tacrolimus, cyclosporine, methotrexate) — TRAE OR vs placeboAdverse0.85 (95% CrI 0.25–2.85), SUCRA 0.690
Notes on safety dataAdverseTRAE analysis based on 18 studies (1,009 treatment arm patients, 880 placebo arm patients); not all trials reported TRAEs; batoclimab phase 3 disproportionately drove FcRn TRAE signal

Subgroup Analysis

Prespecified sensitivity/covariate analyses: (1) Meta-regression for age, sex, and baseline QMG/MG-ADL scores demonstrated confounding of QMG effects by age (β=2.09; 0.78–3.27) and sex (β=−3.16; −4.45 to −1.76). (2) Restricted to AChR-antibody positive patients (19 studies for QMG, 15 for MG-ADL): FcRn largest effect. (3) MuSK-antibody positive subgroup: rozanolixizumab 7 mg/kg had greatest MG-ADL reduction. (4) Phase 3–only analysis: FcRn inhibitors QMG −4.03; C5i −2.42; CD19+ BCDT −2.5 (still comparable and not clinically meaningfully different). (5) 52-week CD19+ BCDT extension analysis: QMG −4.3, MG-ADL −2.79 — largest of any treatment when durability accounted for. (6) CD20+ BCDT (rituximab) QMG effect increased after adjusting for age and sex, likely reflecting the atypical older, male-predominant RINOMAX/BeatMG populations.


Criticisms

  • Excluded pivotal azathioprine and thymectomy RCTs and one 2008 international phase 3 mycophenolate mofetil trial because they reported medians or did not report change from baseline — limits comparison with foundational MG therapies.
  • Meta-analysis of aggregate data, not individual-patient data — precluded sensitivity analyses for disease duration, early-onset vs late-onset MG, refractoriness, or thymectomy status.
  • Covariate-adjusted treatment effects are model-based predictions and should not be interpreted as Class I subgroup evidence.
  • Bayesian hierarchical models limited to simple linear meta-regression due to computational constraints; more complex interactions were not modeled.
  • TRAE reporting was inconsistent across trials — only 18 of 27 trials contributed to safety analysis; batoclimab phase 3 disproportionately drove the FcRn TRAE signal, and IVIg TRAE OR came from a single trial.
  • Heterogeneity in trial length (4–52 weeks), assessment timepoints, and mechanism-specific temporal response patterns (e.g., cyclical FcRn effect vs sustained CD19+ BCDT benefit) complicate direct efficacy comparison.
  • For some studies, MG-ADL/QMG data had to be extracted from published figures using an online tool (graphreader) — introduces potential measurement error, notably for efgartigimod phase 2/3 week-4 data.
  • Several IST trials (2 MMF, 2 cyclosporine, 1 tacrolimus, 2 batoclimab) were single-country studies, limiting generalizability.

Funding

Not explicitly reported in the article body; study was investigator-initiated (Yale School of Medicine) with PROSPERO registration CRD42023428733.

Based on: MG-BNMA (Neurology, 2026)

Authors: McLaren NP, Rosati M, Zhong W, ..., Roy B

Citation: Neurology. 2026 Aug 11;107(3):e218351.

Content summarized and formatted by NeuroTrials.ai.