Diagnostic performance of the 2023 International MOGAD Panel criteria in broad diagnostic-evaluation and MOG-IgG-positive enriched cohorts: a systematic review and meta-analysis.

BACKGROUND: Validation studies have applied the 2023 International MOGAD Panel criteria at different points in the diagnostic pathway. We evaluated broad diagnostic-evaluation and MOG-IgG-positive enriched cohorts separately. METHODS: We searched four databases on 8 July 2026, updated them on 15 July 2026, and archived ClinicalTrials.gov on 16 July 2026. Eligible studies evaluated the 2023 framework against a reference diagnosis; those providing 2 × 2 classifications entered quantitative synthesis. Two reviewers independently selected studies, extracted data, and applied QUADAS-2. The co-primary bivariate random-effects syntheses were defined by pathway without combining overlapping cohorts. RESULTS: Fifteen reports from 14 datasets were included; 10 cohorts (2397 participants) provided 2 × 2 data and nine non-overlapping cohorts (2334 participants) entered quantitative synthesis. In three broad cohorts, sensitivity was 98.6% (95% CI 94.2-99.7%; prediction interval 87.4-99.9%) and specificity was 99.2% (95% CI 97.9-99.7%; prediction interval 97.7-99.8%). In six MOG-IgG-positive enriched cohorts, sensitivity was 94.3% (95% CI 90.4-96.7%; prediction interval 84.4-98.1%) and specificity was 63.6% (95% CI 44.7-79.0%; prediction interval 25.7-89.8%). Seven of 10 cohorts were at high risk of reference-standard bias; certainty was very low. CONCLUSIONS: The two pathways answer different diagnostic questions and should not be represented by one transportable accuracy estimate. Sparse broad-cohort evidence suggested high accuracy, whereas specificity was lower and less certain among seropositive patients. Retrospective application, incorporation bias, sparse reference-negative groups, and tertiary-centre sampling limit generalisability. TRIAL REGISTRY: PROSPERO registration: CRD420261446057.