Prediction of relapse in myelin oligodendrocyte glycoprotein antibody-associated disease: external validation of the MOG-AR score

TL;DR

This study tested a relapse prediction score for MOGAD in a UK cohort and found it performed poorly, suggesting the need for better tools to guide early treatment decisions.

Contribution

The study externally validates the MOG-AR score in a UK cohort, revealing its suboptimal performance for predicting relapses in MOGAD.

Findings

38% of MOGAD patients relapsed within 3 years of onset in the UK cohort.

MOG-AR score showed suboptimal discrimination with an AUC of 0.58 for predicting relapse.

Calibration indicated overestimation of relapse probabilities by the MOG-AR score.

Abstract

Predicting relapses in myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) when disability is relapse-dependent is crucial to guide treatment decisions including whether to treat from onset. MOG-AR, a relapse risk score, was recently developed in a Chinese cohort from disease onset. This study aimed to externally validate the MOG-AR score.

MOGAD patients seen through the Oxford National NMO Service with ≥ 1-year disease duration and available data for MOG-AR score calculation (variables of age, sex, onset attack phenotype, treatment) were included. MOG-AR score and grade were calculated. Relapse occurrence at 3 years from onset was used as the primary outcome. MOG-AR performance was assessed by measures of discrimination and calibration.

We included 284 MOGAD patients with a 4.7-year median disease duration. Relapse occurred in 38% within 3 years of onset. Median MOG-AR score and grade were 11 (IQR 10–12) and 3 (3–3) for those who relapsed and 9 (8–11) and 3 (2–3) for those who did not. Observed proportion with relapse was 27%, 26%, 41%, and 53% for grades 1, 2, 3, and 4, respectively. Discrimination assessment by grade showed an area under the receiver operating characteristic curve of 0.58 (95% CI 0.52–0.63). Calibration assessment was consistent with overestimation of relapse probabilities.

In a UK-MOGAD cohort, MOG-AR score showed suboptimal performance in predicting relapse over a 3-year period from onset. Further work to find clinically accessible predictive biomarkers and tools for a relapsing course would facilitate better treatment strategies near disease onset.

Figures