Predicting relapses in MOGAD