Long-Term Clinical and Biological Prognostic Factors of Anti-NMDA Receptor Encephalitis in Children.

Abstract

Background and objectives: Anti-NMDAR encephalitis (NMDARE) is a severe neurologic condition, and recently, the NMDAR Encephalitis One-Year Functional Status (NEOS) score has emerged as a 1-year prognostic tool. This study aimed to evaluate NEOS score and biomarker (neurofilament light chains [NfL], total-Tau protein, glial fibrillary acidic protein, and serum cytokines) correlation with modified Rankin Scale (mRS), cognitive impairment, and clinical recovery in pediatric NMDARE over 2 years.

Methods: In this French multicenter observational study, 104 pediatric patients with NMDARE were followed for a minimum of 2 years. Clinical data and serum/plasma samples were collected. Biomarker levels, measured using electroluminescence mesoscale discovery (MSD) S-PLEX, were compared between patients and controls and assessed for correlations with disease activity, mRS, cognitive/language impairment, and recovery status at 2 years.

Results: At a median follow-up of 39.5 months, 68 percent of patients had unfavorable recovery and 54% had significant cognitive impairment. Both outcomes were strongly associated with younger age at diagnosis (OR 6.10 [1.91-27.3] p < 0.01 and 5.69 [1.46-27.7] p = 0.02, respectively). A higher NEOS score was significantly correlated with increased cognitive impairment (OR 2.53 [1.52-4.21], p < 0.001), higher mRS scores (OR 2.12 [1.34-3.57], p < 0.01), and unfavorable recovery at 2 years (OR 2.00 [1.30-3.06], p = 0.015). Elevated NfL levels were significantly associated with unfavorable recovery (OR 3.62 [1.29-10.9] p = 0.012) and severe cognitive impairment (OR 3.77 [1.38-10.9] p = 0.012) at 2 years. The combined area under the curve (AUC) for NfL and NEOS was significantly higher than the AUCs of NEOS and NfL alone (p = 0.01).

Discussion: The NEOS score strongly predicts long-term outcomes in NMDARE, with its predictive value extending beyond the first-year mR prediction. NfL levels at disease onset seem to improve accuracy in predicting poor outcomes, providing valuable information for treatment decisions and future clinical trials.