Abstract
OBJECTIVE: This retrospective observational study primarily aimed to analyse the clinical characteristics of patients with neuronal surface antibody-mediated autoimmune encephalitis (AE) in China and report their prognosis after immunotherapy. METHODS: Clinical characteristics, laboratory or imaging examinations, and treatment outcomes of 103 patients diagnosed with AE between 1 September 2014 and 31 December 2020 were collected. Univariate and multivariate logistic regression analyses were performed to determine the predictors of poor prognosis. RESULTS: Overall, 103 patients were enrolled in the study. The main clinical symptoms included seizures (74.8%), psychiatric and behavior disorders (66.0%), cognitive deficits (51.5%), disturbances of consciousness (45.6%), and movement disorders/involuntary movements (26.2%). The distribution of clinical syndromes also differed for different AE subtypes. The efficacy rates of first-line immunotherapy for anti-NMDAR, anti-LGI1, anti-GABA(B)R, and anti-CASPR2 encephalitis were 70.2%, 92.3%, 70%, and 83.3%, respectively, and rituximab was administered to 21 patients as second-line immunotherapy, including 14 patients with anti-NMDAR encephalitis, 4 with anti-LGI1 encephalitis, 2 with anti-GABA(B)R encephalitis, and 1 with anti-CASPR2 encephalitis. Five patients with poor effect of the second-line treatment received bortezomib. According to the results of the last follow-up, 78 patients had a good prognosis (mRS 0-2), and 21 patients had a poor prognosis (mRS 3-6). The proportion of patients with a poor prognosis was significantly higher in anti-GABA(B)R encephalitis compared to the other AE subtypes (p<0.001). Multivariate analysis indicated that elevated neutrophil-to-lymphocyte ratio (NLR) and tumour presence were independent risk factors for poor prognosis. The regression equation of the model was logit(P)=-3.480 + 0.318 NLR+2.434 with or without tumour (with assignment =1, without assignment =0). The prediction probability generated by the regression model equation was used as the independent variable for receiver operating curve (ROC) analysis. The results showed that the area under the curve (AUC) of the prediction probability was 0.847 (95% CI, 0.733-0.961; p < 0.001). CONCLUSIONS: Different AE subtypes demonstrated different clinical symptom spectra throughout the disease stage. Anti-LGI1 encephalitis and anti-CASPR2 encephalitis were more sensitive to first-line and second-line treatments. Anti-GABA(B)R encephalitis had the worst prognosis among the abovementioned subtypes. The regression equation constructed using NLR and tumour presence effectively predicted the poor prognosis.