Introduction:An important challenge in diagnosing anti-NMDAR encephalitis (NMDARe) is differentiating it from a first episode of psychosis (FEP) caused by a psychiatric disease (pFEP). CSF antibody testing distinguishes these diseases, but spinal taps are difficult to obtain in psychiatric facilities. A separate problem is the lack of biomarkers of NMDARe severity and outcome. Here we assessed the performance of neurofilament light chain (NfL) testing in these settings.Methods:In this observational study, NfL were determined with Single molecule array (SiMoA) in patients with NMDARe, pFEP, herpes simplex encephalitis (HSE), and healthy subjects (HC), the latter two groups used as controls. Receiver operating characteristic (ROC) analyses were performed to assess the prediction accuracy of serum NfL (sNfL) levels for NMDARe and pFEP, and to obtain clinically useful cutoffs.Results:118 patients with NMDARe (33 with isolated psychosis at presentation), 45 pFEP, 36 HSE, and 36 HC were studied. NMDARe patients with seizures/status epilepticus, ICU admission, CSF pleocytosis (>20 WBC/µL), and without early immunotherapy were more likely to have higher sNfL than NMDARe without these features. NfL levels at diagnosis of NMDARe did not correlate with outcome at 1 year follow-up assessed with the modified Ranking Scale (mRS). NMDARe patients had significantly higher NfL than pFEP and HC, and lower than HSE patients. ROC analysis of sNfL between NMDARe with isolated psychosis and pFEP provided an AUC of 0.93 (95% CI 0.87-0.99) and a sNfL cutoff ≥15 pg/mL to distinguish these disorders (sensitivity 85%, specificity 96%, positive likelihood ratio 19.3). Forty-three/45 (96%) pFEP had sNfL<15pg/mL whereas only 5/33 (15%) NMDARe with isolated psychosis were below this cutoff (risk estimation NMDARe vs pFEP: odds ratio 120.4 [95% CI 21.8-664], p<0.001). None of the HSE and 35/36 (97%) HC had sNfL<15pg/mL.Discussion:NfL measured at diagnosis of NMDARe associated with features of disease severity but not with long-term outcome. Young patients with FEP and sNfL≥15pg/mL had 120 times higher chance of having NMDARe than pFEP. This cutoff correctly classified 96% of pFEP and 85% of NMDARe with isolated psychosis. Patients with FEP of unclear etiology and sNfL≥15pg/mL should undergo CSF NMDAR-antibody testing.