Presentation (CTI)

OBJECTIVES: Corticosteroid burden in myasthenia gravis (MG) is substantial, but novel biologic therapies offer the potential to reduce steroid dependence. This study sought to characterise real-world corticosteroid prescribing patterns and quantify longitudinal steroid exposure among MG patients treated with novel therapies.
METHODS: Electronic medical records (EMRs) from the Vitaccess Real MG (VRMG) registry were analysed to characterise corticosteroid prescribing patterns among patients receiving novel MG therapies including complement, FcRn and B-cell targeting agents. Corticosteroid (prednisolone/prednisone/methylprednisolone) prescriptions and dosing patterns were extracted using clinical codes and a pattern-recognition algorithm, enabling detailed reconstruction of longitudinal treatment trajectories. Doses were converted to cumulative prednisolone-equivalent units (mg).
RESULTS: As of June 2026, VRMG included 189 respondents. 109 recruited via direct-to-patient pathways had linked EMR medication data. Among these, 32 (29.4%) received prescriptions for at least one novel MG therapy, with 30 (93.7%) concurrently prescribed corticosteroids. In total, 645 steroid prescriptions were recorded (median 15, mean 21.5 per participant), including 53 titration regimens, 220 single-course prescriptions and 144 “as-needed” prescriptions, reflecting highly heterogeneous, complex and episodic use patterns. 96.9% of prescriptions were successfully converted to dose equivalents: 20 records were not converted (14 steroid eye drops, 5 suspensions, 1 tablet formulation). Among 15 participants with at least 12 months of longitudinal follow-up before and after initiation of their most recent novel therapy, median cumulative annual steroid exposure was lower after novel treatment initiation (3,029 mg) compared with the preceding year (4,640 mg).
CONCLUSIONS: EMR-derived data can provide granular, longitudinal measures of corticosteroid burden in MG despite dynamic and heterogeneous prescribing patterns. These findings demonstrate the feasibility and value of converting dosing data into decision-relevant burden measures and provide an early real-world signal of lower steroid exposure following novel treatment initiation. Ongoing refinement incorporating natural language processing may improve precision of steroid burden measurement in complex regimens.