← Back
NeuroTrials.ai
Neurology Clinical Trial Database

MOGAD-PLEX

Effectiveness and Safety of Plasmapheresis in MOGAD Attacks: A Systematic Review and Meta-Analysis

Year of Publication: 2026

Authors: Vilardo M, Mahler JV, Leles Vieira de Souza B, ..., Matiello M

Journal: Neurology

Citation: Neurology. 2026 Aug 11;107(3):e218300. Published online 2026 Jul 22.

Link: https://doi.org/10.1212/WNL.0000000000218300

Bottom Line

In this systematic review and meta-analysis of 10 observational studies (1,045 patients, 1,368 attacks; >120 centers, 14 countries), plasmapheresis was associated with clinically meaningful visual and neurologic recovery in acute MOGAD attacks — particularly in isolated optic neuritis (excellent recovery 81%) and steroid-refractory presentations (logMAR mean difference 1.22; I²=0%) — with a low adverse-event rate (3%) and rare serious events (0.5%); however, the evidence is low-to-very-low certainty (GRADE), all included studies are observational with moderate risk of bias, and randomized trials are needed to define patient selection and timing.

Major Points

  • Systematic review and meta-analysis (PROSPERO CRD420251169201) searching PubMed, Embase, Web of Science, and Cochrane Library from inception through October 20, 2025; PRISMA 2020 methodology; 2,744 records screened → 10 studies (n=1,045 patients / 1,368 attacks) in qualitative synthesis; 8 studies (n=710 / 1,033 attacks) in the meta-analysis.
  • All included studies observational (7 retrospective, 3 prospective); multicenter in 7, single-center in 3; representing >120 centers across 14 countries; publications 2016–2025; 4 studies (65.6% of attacks) used the 2023 MOGAD diagnostic criteria.
  • Baseline: mean age 35 ± 17 years, 56.5% female; attack manifestations dominated by optic neuritis (41%) and combined ON+myelitis±cerebral (29.4%); mean time from clinical onset to plasmapheresis 15.2 ± 13 days; PLEX regimens median 5 sessions (IQR 5–7, range 3–11) on alternate days.
  • Primary efficacy — visual acuity (logMAR) pre vs post-PLEX: mean difference 1.43 (95% CI 0.63–2.24; Q=19.38, p<0.0001; I²=90%; 3 studies; 341 optic neuritis attacks). In steroid-refractory subgroup (PLEX after IVMP), mean difference 1.22 (95% CI 0.92–1.52; Q=0.12, p=0.73; I²=0%; 2 studies; 249 attacks) — heterogeneity resolved.
  • Excellent recovery: plasmapheresis 47% (95% CI 23–71%; Q=730.56, p<0.0001; I²=99%; 10 study groups; 244/491 attacks); IVMP alone 76% (2–100%; I²=98%; 3 groups; 120/211). PLEX-after-IVMP subanalysis: 29% (14–45%; I²=84%; 6 groups; 134/366). Isolated optic neuritis PLEX excellent recovery 81% (33–100%; I²=79%; 4 groups; 109/116 attacks) at 3–6 mo follow-up.
  • Good recovery: plasmapheresis 94% (95% CI 88–100%; Q=26.30, p=0.0004; I²=73%; 8 groups; 439/473 attacks); IVMP alone 73% (21–100%; I²=96%; 5 groups; 143/182). PLEX-after-IVMP subanalysis 81% (60–100%; I²=78%; 6 groups; 120/334).
  • Safety: plasmapheresis-related adverse events 3% (95% CI 1–14%; I²=0%; 4 studies; 24/444 patients); serious AEs 0.5% (95% CI 0–2%; 2/444). Reported complications included transient rash, fresh frozen plasma reaction (1 patient, PLEX stopped after 4 cycles), and internal jugular vein thrombosis leading to premature PLEX termination in a single case (Schwake 2024).
  • Meta-regression showed no clear association between baseline logMAR severity and magnitude of improvement (slope 0.10 per unit baseline; 95% CI −10.07 to 10.26; p=0.92) — but result was highly imprecise given few studies. Publication bias assessment by funnel plot; Egger regression not performed (<10 studies).
  • Quality: all studies classified as moderate risk of bias by ROBINS-I (confounding [D1] and selection of reported results [D7] were dominant domains). GRADE overall certainty of evidence for both effectiveness and safety was low to very low.

Design

Study Type: Systematic Review and Random-Effects Meta-Analysis of Observational Studies (PROSPERO CRD420251169201)

Randomization:

Blinding: N/A (meta-analysis of observational studies; no randomized trials included)

Enrollment Period: Search from database inception through October 20, 2025; included studies published 2016–2025

Follow-up Duration: Overall 11 ± 12.2 months (n=836); primary outcome at last follow-up (mean 6 ± 2.2 months); individual studies ranged from 3 months to 43 ± 30 months

Centers: >120 centers across included studies (7 multicenter, 3 single-center)

Countries: 14 countries across the included studies (multicontinental)

Sample Size: 1045

Analysis: R v4.3.3 with meta package v7.0-0. Continuous outcomes (logMAR): unpaired inverse-variance random-effects mean difference; paired sensitivity analyses. Proportions (excellent/good recovery): random-effects raw proportions (PRAW) with REML and Knapp-Hartung; secondary Freeman-Tukey double-arcsine transformations. Adverse events: generalized linear mixed model (GLMM). Heterogeneity: Cochran Q, I² (0–40% low, 40–60% moderate, >60% high), leave-one-out, and Baujat analyses. Multiarm studies: independent-arm approach with sensitivity analyses collapsing arms. Publication bias: funnel plot (Egger regression not performed given <10 studies). Meta-regression on baseline logMAR. False discovery rate 5% via Benjamini-Hochberg. Quality: ROBINS-I for observational studies; GRADE for certainty of evidence.


Inclusion Criteria

  • Studies with at least 5 patients (any language) published from database inception through October 20, 2025
  • Diagnosis per 2023 MOGAD international diagnostic criteria (Banwell et al.) — OR (for pre-2023 studies) MOG-IgG positive in serum by cell-based assay with an inaugural MOGAD-typical syndrome
  • Treatment with intravenous methylprednisolone (IVMP), intravenous immunoglobulin (IVIG), or plasmapheresis (plasma exchange or immunoadsorption) for an acute MOGAD attack
  • Reported ≥1 outcome of interest (visual acuity by Snellen/logMAR, excellent/good recovery, plasmapheresis-related adverse events, or EDSS/mRS)
  • No age restriction — both adult and pediatric patients eligible
  • Included both prospective and retrospective observational cohort designs (RCT eligible but none identified)

Exclusion Criteria

  • Absence of ≥1 outcome of interest
  • Case reports (<5 patients)
  • Conference abstracts
  • Editorials and commentaries
  • Studies of MOG-IgG-positive individuals without a MOGAD-typical clinical syndrome (to avoid low-titer false-positive misclassification)
  • Gray-literature / unpublished reports (excluded due to insufficient data for quality assessment)

Baseline Characteristics

CharacteristicAll patients (systematic review, 10 studies)Meta-analysis subset (8 studies)Included studies (10 total)
Patients (n)1,045710
Attacks (n)1,3681,033
Female591 (56.5%)420 (59.1%)
Male454 (43.5%)290 (40.9%)
Age, years (mean ± SD)35 ± 1736 ± 17
Disease duration, years (mean ± SD)3.77 ± 5.79 (n = 315)
Time from clinical onset to IVMP, days (mean ± SD)12 ± 8.8 (n = 70)
Time from clinical onset to plasmapheresis, days (mean ± SD)15.2 ± 13 (n = 359)
Clinical follow-up, months (mean ± SD)11 ± 12.2 (n = 836)6 ± 2.2 (n = 501)
Overall EDSS at presentation (mean ± SD)3.5 ± 2.3 (n = 511)4 ± 2 (n = 271)
Overall logMAR at presentation (mean ± SD)1.56 ± 0.6 (n = 436) — Snellen equivalent ~20/700
Attack manifestationsOptic neuritis 545 (41%) [unilateral 152, bilateral 72, not specified 321]; combined ON+myelitis±cerebral 391 (29.4%); myelitis 162 (12.2%); ADEM 85 (6.4%); cortical encephalitis 39 (2.9%); multifocal 39 (2.9%); brainstem/cerebellar 37 (2.8%); cerebral deficits alone 21 (1.6%); other 7 (0.5%)Optic neuritis 400 (40.2%); combined ON+myelitis±cerebral 391 (39.3%); myelitis 112 (11.3%); multifocal 39 (3.9%); ADEM 24 (2.4%); brainstem/cerebellar 21 (2.1%); cerebral deficits alone 4 (0.4%); cortical encephalitis 3 (0.3%)
Acute treatment (of 1,319 attacks with data)IVMP alone 701 (53.2%); IVMP then PLEX 338 (25.6%); IVMP prior or simultaneously with PLEX 90 (6.8%); no treatment 70 (5.3%); IVMP and IVIG then PLEX 35 (2.7%); PLEX/IA alone 29 (2.2%); IVMP with IVIG 28 (2.1%); IVMP then IA 25 (1.9%); IVIG alone 3 (0.2%)
Plasmapheresis regimenMedian 5 sessions (IQR 5–7, range 3–11) on alternate days
EDSS pre vs post-IVMP (mean ± SD, n=20)4.6 ± 2.18 vs 0.87 ± 1.18 (Devi 2025, Kosior 2024)
EDSS pre vs post-plasmapheresis (mean ± SD, n=243)4.5 ± 2.59 vs 1.17 ± 1.85 (Thakolwiboon 2025)
logMAR pre vs post-IVMP (n=18)0.73 ± 0.53 → 0.16 ± 0.06
logMAR pre vs post-plasmapheresis (n=341)1.42 ± 1.06 → 0.07 ± 0.35
Plasmapheresis-related adverse events24 / 444 (~5% crude)
Serious plasmapheresis-related adverse events2 / 444 (~0.5% crude)
Acute treatment (of 1,006 attacks with data)IVMP alone 487 (48.4%); IVMP then PLEX 326 (32.4%); IVMP prior or simultaneously with PLEX 90 (8.9%); IVMP and IVIG then PLEX 35 (3.5%); PLEX/IA alone 29 (2.9%); IVMP then IA 25 (2.5%); no treatment 14 (1.4%)
Chen et al., 2023 (Am J Ophthalmol)International multicenter, 395 optic neuritis attacks; PLEX after or simultaneously with IVMP; final VA at 6–12 mo
Devi et al., 2025 (Cureus)Prospective observational; NMOSD/MOGAD comparison; IVMP outcomes reported; f/u at 1, 3, 6, 12 mo
Fu et al., 2022 (Neurol Ther)Prospective cohort; PLEX after IVMP for demyelinating ON; logMAR at 7 time points to 6 mo
Jarius et al., 2016 (J Neuroinflammation)MOG-IgG-positive multicenter; PLEX/IA after IVMP or alone; mean f/u 75 ± 46.5 mo
Kosior et al., 2024 (Mult Scler Relat Disord)MOGAD/NMOSD; IVMP and no-treatment groups; f/u ≥3 mo
Rode et al., 2022 (JNNP)Inaugural adult MOG-IgG+ ON; IVMP alone vs IVMP + PLEX; f/u 3 mo
Schwake et al., 2024 (JNNP)117 apheresis interventions in 571 attacks; PLEX/IA after IVMP
Thakolwiboon et al., 2025 (Neurology)PLEX for MOGAD; EDSS/VA pre and post; median f/u 7 mo (4–11); early subsequent attack rate 6% within 3 mo
Armangue et al., 2020 (Lancet Neurol) [SR only]Pediatric demyelinating/encephalitic syndromes with MOG-IgG; mRS at 43 ± 30 mo
Kwon et al., 2024 (JAMA Neurol) [SR only]Time-to-treat first acute MOGAD attack; EDSS and ARR at ≥12 mo

Arms

FieldPlasmapheresis-containing regimens (PLEX or immunoadsorption)Control
InterventionPlasmapheresis (PLEX or IA), median 5 sessions (IQR 5–7, range 3–11) on alternate days — given as escalation after IVMP, simultaneously with IVMP, following IVMP + IVIG, or as monotherapy for acute MOGAD attacksHigh-dose intravenous methylprednisolone monotherapy for acute MOGAD attacks (5 studies contributed IVMP-only outcomes)
DurationMedian 5 sessions; outcomes assessed at last follow-up (mean 6 ± 2.2 months in meta-analysis subset)Matched follow-up per included study

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Change in visual acuity measured as logMAR (pretreatment minus follow-up) after plasmapheresis in patients with optic neuritis — random-effects unpaired inverse-variance mean differencePrimaryIVMP alone (comparison not statistically computable due to lack of baseline + follow-up logMAR in IVMP-only cohorts)Plasmapheresis (PLEX or IA)<0.0001
logMAR change (steroid-refractory subgroup: PLEX after IVMP)Secondary0.73 (Q=0.12)
Excellent recovery — plasmapheresis (any presentation)Secondary<0.0001 (Q=730.56)
Excellent recovery — IVMP aloneSecondary<0.0001 (Q=114.24)
Excellent recovery — plasmapheresis after IVMP subanalysisSecondary<0.0001 (Q=31.86)
Excellent recovery — isolated optic neuritis after plasmapheresis (3–6 mo)Secondary0.002 (Q=14.17)
Excellent recovery — isolated optic neuritis after plasmapheresis (≥6 mo)Secondary<0.001 (Q=11.58)
Poor recovery — isolated optic neuritis after plasmapheresis (≥6 mo)Secondary<0.001
Good recovery — plasmapheresis (any presentation, ON + myelitis + cerebral)Secondary0.0004 (Q=26.30)
Good recovery — IVMP aloneSecondary<0.0001 (Q=91.69)
Good recovery — plasmapheresis after IVMP subanalysisSecondary0.0003 (Q=69.88)
Meta-regression: baseline logMAR vs magnitude of improvementSecondary0.92 (NS; imprecise given few studies)
Plasmapheresis-related adverse events (any)Adverse3% (95% CI 1–14%; Q=0.00, p=0.99; I²=0%; 4 studies [Chen 2023, Fu 2022, Schwake 2024, Thakolwiboon 2025]; 24/444 patients)
Serious plasmapheresis-related adverse eventsAdverse0.5% (95% CI 0–2%; Q=0.23, p=0.97; I²=0%; 4 studies; 2/444 patients)
Reported individual complicationsAdverseTransient rash (Fu 2022, well tolerated without discontinuation); fresh frozen plasma reaction leading to PLEX discontinuation after 4 cycles in 1 patient (visual acuity improved to 20/25 at that time; Thakolwiboon 2025); internal jugular vein thrombosis leading to premature apheresis termination in 1 patient (Schwake 2024); Schwake 2024 reported 8 procedural complications (7%): 5 with PLEX, 3 with combined PLEX/IA; Thakolwiboon 2025 reported 16 events (7%)
Serious adverse events describedAdverseCentral line placement complications (e.g., pneumothorax, hemothorax, sepsis) requiring invasive intervention or therapy discontinuation — occurred rarely
Non-serious AEs typically monitoredAdverseHemodynamic effects (hypotension), electrolyte/coagulation abnormalities, infusion-related reactions, systemic/local symptoms — all managed successfully in retrospective series
Reporting quality noteAdverseNon-serious AEs are often incompletely reported in retrospective cohorts; safety estimates should be interpreted with caution

Subgroup Analysis

Prespecified subgroups: (1) steroid-refractory patients (PLEX after incomplete IVMP response) — heterogeneity for logMAR change resolved (I²=0%; MD 1.22, 95% CI 0.92–1.52) and PLEX-after-IVMP good recovery 81%; (2) isolated optic neuritis — excellent recovery 81% (33–100%) at 3–6 mo, 69% at ≥6 mo. Data insufficient to analyze pooled EDSS scores or isolated myelitis presentation separately. Leave-one-out: omission of Chen 2023 or Thakolwiboon 2025 resolved logMAR heterogeneity (I²=0%); heterogeneity for excellent recovery remained high across all analyses in PLEX group; in IVMP-alone group, omitting Jarius 2016 reduced I² to 37%; good recovery — omitting Chen 2023 in PLEX group reduced I² to 53%; omitting myelitis subgroup in Devi 2025 in IVMP-alone group reduced I² to 0%. All Benjamini-Hochberg-adjusted p-values <0.01; meta-regression on baseline logMAR remained NS after FDR correction.


Criticisms

  • All 10 included studies were observational (7 retrospective, 3 prospective); no randomized trials of plasmapheresis in MOGAD exist — findings should be considered hypothesis-generating rather than definitive.
  • GRADE certainty of evidence for both effectiveness and safety was low to very low; ROBINS-I rated all studies at moderate risk of bias, dominated by confounding (D1) and selection of reported results (D7).
  • Extreme heterogeneity across pooled outcomes (I² 73–99% for most efficacy analyses), reflecting variability in study design, populations, criteria for diagnosis/relapse/recovery, treatment timing, dosing, sequencing, phenotype (ON vs myelitis vs cerebral), and baseline severity.
  • Indication bias / confounding by severity: patients selected for plasmapheresis had more severe or steroid-refractory attacks than IVMP-only cohorts, biasing direct comparisons and paradoxically resulting in a lower pooled 'excellent recovery' rate for PLEX (47%) than for IVMP-alone (76%).
  • Meta-analysis relies on indirect comparisons across cohorts rather than head-to-head randomized data; transitivity and consistency assumptions may not be satisfied.
  • Only 3 studies contributed to the primary logMAR analysis (n=341 optic neuritis attacks); comparison group logMAR was not statistically computable due to absent baseline + follow-up VA measurements in IVMP-only cohorts.
  • EDSS scores were insufficiently reported for pooled analysis; medians/IQRs were approximated to means/SDs in 2 studies, imprecise for skewed data.
  • Data insufficient to formally analyze the isolated myelitis phenotype or to compare plasmapheresis vs IVIG; also insufficient granularity for the most clinically informative subgroups (severe baseline VA, IVMP failure with treatment ≤2 weeks of onset, ≥4 PLEX sessions).
  • Definitions of 'steroid-refractory' varied substantially across studies (typically VA ≤20/200, <3-line Snellen improvement, or failure to regain function), complicating cross-study synthesis.
  • Non-serious adverse events are known to be incompletely reported in retrospective studies — safety estimate (3%) may be an underestimate.
  • Publication bias assessed only by funnel plot (Egger regression not performed given <10 studies per outcome); gray literature was excluded.
  • Real-world generalizability limited by heterogeneous demographic composition (ages, ethnicities) and reliance on centers with expertise in MOGAD care.

Funding

The authors report no targeted funding. The authors report no relevant disclosures (full disclosures available at Neurology.org/N).

Based on: MOGAD-PLEX (Neurology, 2026)

Authors: Vilardo M, Mahler JV, Leles Vieira de Souza B, ..., Matiello M

Citation: Neurology. 2026 Aug 11;107(3):e218300. Published online 2026 Jul 22.

Content summarized and formatted by NeuroTrials.ai.