Predicting Functional Outcome and Response to Therapy of Anti-NMDAR Encephalitis, Already at Diagnosis: The NEOS2 Model.
Juliette Brenner1, Anna Bastiaansen1, Marienke de Bruijn3, Frank Leypoldt4, Jerome Honnorat5, Josep Dalmau6, Takahiro Iizuka7, Mareike Jansen4, Ina Schroder4, Sergio Muniz-Castrillo8, Yvette Crijnen1, Juna de Vries1, Marco Schreurs2, Amaia Munoz6, Mar Guasp6, Peter Sillevis Smitt1, Maarten Titulaer1
1Neurology, 2Immunology, Erasmus University Medical Center, 3Neurology, St. Elisabeth Hospital, 4University Kiel, 5Hospices Civils de Lyon, 6Hospital ClĂ­nic Barcelona - IDIBAPS, 7Neurology, Kitasato University School of Medicine, 8Stanford university
Objective:

Develop and validate a model to predict functional outcome and response to first-line therapy at the time of diagnosis of anti-NMDAR encephalitis (anti-NMDARE).

Background:
Anti-NMDARE is a severe, but treatable neurological condition, with considerable and variable long-term disability. The available NEOS score predicts outcome a month into treatment. To predict outcome and response to immunotherapy at the time of diagnosis would be a big improvement. This would timely identify patients needing aggressive treatment and avoid harmful side-effects in those with good outcome.
Design/Methods:

Data of five anti-NMDARE cohorts (the Netherlands, Germany [GENERATE], France, Spain and Japan) were combined. The datasets per country will be randomly split for development (70%) and validation (30%).

The primary outcome is functioning at one year, measured by the modified Rankin Scale (mRS). Secondary outcomes are effect of first-line therapy and ability to go back to work or school.

We have performed preliminary logistic regression analyses.

Results:
We have included 720 patients (79% female; mean age 25 years, 95% Confidence-Interval [CI] 2-67 years). 79% of patients had good functionality after one year, 74% was able to return to work/school. In univariable analysis, twelve predictive variables for functional outcome were identified. In a preliminary multivariable analysis, five independent predictive variables remained significant: age at disease onset, diagnostic delay, dysautonomia, ICU admission and CSF leukocyte count. A very preliminary five-point model, accurately classifies patients into outcome categories (97% good outcome if 0 points to 27% if 5 points). The model also seems to predict response to first-line therapy (from 87% response if 0 points to 27% if 5 points). We are currently improving the model.
Conclusions:

In anti-NMDAR encephalitis, outcome at one year and response to first-line therapy can already be predicted at the time of diagnosis.

10.1212/WNL.0000000000203481