This retrospective study was approved by the institutional review boards of the Zhejiang Provincial People’s Hospital and the Sir Run Run Shaw Hospital of Zhejiang University. Written informed consent was obtained from all patients. We reviewed the consecutive medical charts of patients from the Zhejiang Provincial People’s Hospital and the Sir Run Run Shaw Hospital of Zhejiang University between October 2016 and December 2023. Diagnosis of AE was established using the diagnostic criteria established in 2016 (2). Serum and CSF samples were tested by cell-based assays for neuronal surface antibodies: NMDAR, LGI1, GABABR, GABA A receptor (GABAAR), contactin-associated protein-like 2 (CASPR2), a-amino-3-hydroxy-5-methyl-4-isoxazolepropinic acid receptor (AMPAR), dipeptidyl peptidase-like protein 6 (DPPX), IgLON5, metabotropic glutamate receptor 5 (mGluR5) and glycine receptor (GlyR). The exclusion criteria were as follows (1): patients with confounding conditions that could potentially affect white blood cell counts, including active infections, hematological disorders, and other systemic autoimmune diseases, were excluded (2). patients with alternative causes of encephalitis/encephalopathy, such as infectious, metabolic, endocrine, psychiatric, or rheumatologic diseases, were also excluded.

Age, sex, age at epilepsy onset, seizure frequency, cognitive symptoms, autoimmune comorbidities, autonomic dysfunction, electroencephalographic (EEG) findings, cerebral magnetic resonance imaging (MRI) findings, cerebrospinal fluid findings, and type of immunotherapy (steroid, intravenous immunoglobulin, rituximab, etc.) were collected.

Venous blood samples were routinely collected for full blood count analysis within 24 hours after hospital admission and before discharge. NLR was defined as the ratio of the neutrophil count to the lymphocyte count. MLR was defined as the ratio of the monocyte count to the lymphocytes count. Neutrophil-to-platelet ratio (NPR) was defined as the ratio of the neutrophil count to the platelet count. Platelet-to-lymphocyte ratio (PLR) was defined as the ratio of the platelet count to the lymphocyte count. SII was defined as the product of platelet and neurophil counts divided by lymphocyte count [(platelet×neutrophil)/lymphocyte].

The participants were regularly followed by neurologists through telephone or in-person interviews conducted every three months. Relapses of AE were defined as recurrence or clear worsening of encephalitis symptoms occurring at least three months after complete remission or stable plateau of prior symptoms, accompanied by deterioration in ancillary testing findings. In the case of relapse or death, the time at which such an event occurs was considered their final follow-up. Based on previous studies (18, 24, 25), the CASE and mRS scores were independently accessed by two neurologists (B.J and S.Z) who were blinded to the diagnosis at the time of admission, discharge and each follow-up. All patients were divided into mild group (CASE ≤ 4) and severe group (CASE ≥ 5) according to the CASE score at admission and discharge. Poor prognosis was described as a CASE score of 5 or higher at the last follow-up visit. Patients who passed away during follow up were assigned a CASE score of 27, indicating maximum severity. Discrepancies in CASE scoring were re-reviewed by two neurologists to determine the final score. If the agreement could not be reached, the case was discussed in a group setting to achieve consensus.

Data were expressed as mean ± standard deviation or median and interquartile ranges (IQR). Categorical variables were analyzed using Chi-square test. If continuous variables were normally distributed, an independent sample t-test or one-way analysis of variance was used. If not, Mann–Whitney U tests or Kruskal–Wallis test was applied. Bonferroni correction was used for multiple comparisons when analyzing clinical characteristics among the three AE subgroups. Statistical significance was set to p < 0.017 (0.05/3). Spearman’s correlation analysis was performed to evaluate the correlation between the CASE score and mRS, the systemic inflammatory factors and disease severity (CASE score). Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive performance for disease severity (CASE score) based on the systemic inflammatory factors. Cut-off values were estimated using the ROC curve, and the corresponding sensitivities and specificities were calculated based on the area under the curve (AUC). Systemic inflammatory factors levels were classified into two groups according to the cut-off values. To identify predictive factors for disease severity and long-term outcomes, variables with p values <0.05 in univariate analysis were included in multivariable logistic regression analyses. P<0.05 indicated statistical significance.

During the study period, a total of 83 patients were enrolled. The median age was 52 ± 19.3 years old, and 52 (62.7%) were male. Among them, 35 (42.2%) patients were diagnosed with anti-NMDAR encephalitis, 24 (28.9%) with anti-LGl1 encephalitis, 13 (15.7%) with anti-GABABR encephalitis, eight with anti-CASPR2 encephalitis, two with anti-IgLON5, anti-DPPX and anti-mGluR5 encephalitis, and one with anti-AMPAR1 encephalitis. In a comparison of the most common subtypes of AE (anti-NMDAR, anti-LGI1, and others), anti-NMDAR encephalitis was found to be more prevalent in younger patients than in other subtypes (P < 0.001). Language problem was more common in anti-NMDAR encephalitis and other AE (P=0.001).

Three patients (one patient with anti-NMDAR encephalitis and two patients with anti-LGI1 encephalitis) were lost to follow-up, and 80 patients remained in the study group. Among 80 patients, 12 patients passed away, six patients experienced a relapse (median 13.5 months, IQR 9.3-32.8 months), and the remaining patients experienced a complete remission or plateauing from prior symptoms (median 36.5 months, IQR 20.0-50.5 months). The baseline clinical characteristics of the patients with different cell surface antibodies mediated AE were summarized in Table 1 .

There was a strong correlation between CASE and mRS scores at admission, discharge and follow-up ( Figure 1 ). The CASE score showed a broader range and more pronounced changes than the mRS within the same patient cohort, suggesting greater sensitivity in capturing disease severity and progression. We further stratified the correlation between the mRS and CASE scores among the three most common subtypes of AE (anti-NMDAR, anti-LGI1, and other AE) at admission, discharge, and follow-up, demonstrating strong correlations within each subgroup ( Figure 1 ). Moreover, the CASE score also demonstrated more pronounced changes when compared to mRS scores across the three subtypes.

Based on the admission CASE scores, 30 patients (36.1%) were classified into the mild group, while 53 patients (63.9%) were placed in the severe group. Spearman’s correlation analysis showed that NLR, MLR, NPR, PLR and SII were positively correlated with admission CASE scores ( Figure 2 ). ROC analysis was performed to assess the predictive value of NLR, MLR, NPR, PLR, and SII for disease severity. The AUC values for NLR, MLR, NPR, PLR, and SII were 0.666, 0.707, 0.631, 0.614, and 0.691, respectively. Details of the optimal cut-off values are provided in Table 2 .

Using univariate analysis, higher levels of NLR, MLR, NPR, PLR, and SII were significantly associated with more severe clinical status at admission ( Table 3 ). To identify independent clinical predictors of disease severity, variables found to be significant in univariate analysis were included in the multivariate logistic regression model. The analysis revealed that psychiatric symptoms, language problem, gait instability and ataxia, and higher level of SII were independent risk factors for severe disease in patients with AE ( Table 3 ). Patients were divided into two groups based on the optimal cut-off value of SII identified through ROC analysis. The corresponding clinical characteristics of the high and low biomarker groups are presented in Supplementary Table 1 .

According to the CASE scores at the time of discharge, the patients were divided into a mild (52 patients, 62.7%) and severe disease (31 patients, 37.3%) groups. No statistically significant correlation was noted between systemic inflammatory factors and CASE score at discharge, using spearman’s correlation analysis. The optimal cut-off values were 3.3, 0.23, 0.05, 176.87 and 1011.05 for NLR, MLR, NPR, PLR, and SII respectively ( Table 2 ).

Univariate analysis revealed higher levels of NLR and PLR were significantly associated with severe disease at the time of discharge ( Table 4 ). Multivariate logistic regression analysis showed that psychiatric symptoms, cognitive dysfunction, language problem, gait instability and ataxia, and higher level of PLR as independent risk factors for disease severity in AE ( Table 4 ).

We further investigated whether systemic inflammatory factors at admission could predict the disease severity at discharge. The NLR, MLR, NPR, and SII at admission were significantly and positively correlated with CASE scores at discharge ( Figure 2 ). The optimal cut-off values were 4.63, 0.62, 0.018, 314.55 and 1923.56 for NLR, MLR, NPR, PLR, and SII respectively ( Table 2 ). Univariate analysis showed higher levels of NLR, MLR, NPR, PLR, and SII at the time of admission were associated with severe disease severity at discharge ( Table 3 ). Multivariate logistic regression analysis revealed that ICU admission and higher level of MLR were independent risk factors for disease severity of AE at discharge ( Table 3 ). Patients were divided into two groups based on the optimal cut-off value of MLR. The clinical characteristics of the two groups are presented in Supplementary Table 2 .

Based on the CASE score, 63 patients (78.8%) were in the mild disease group and 17 patients (21.2%) were in the severe group at the last follow-up. We further investigated whether systemic inflammatory factors at the time of admission could predict the disease severity at follow-up. Spearman’s correlation analysis showed that there was no significant correlation between systemic inflammatory factors at admission and CASE score at follow-up. The AUC values for NLR, MLR, NPR, PLR, and SII were 0.632, 0.616, 0.711, 0.346 and 0.592, respectively ( Table 2 ).

Univariate analysis showed that compared with the good prognosis group, the poor prognosis group had higher elevated levels of NPR, PLR and SII at admission. Multivariate logistic regression analysis showed older age at onset, and higher level of NPR were independent risk factors for the poor prognosis of patients with AE ( Table 3 ). Based on the optimal cut-off value of NPR, two groups of patients were stratified. The clinical features of both groups are outlined in Supplementary Table 3 .

We also investigated whether systemic inflammatory factors at the time of discharge could predict the disease severity at follow-up. No statistically significant correlations of systemic inflammatory factors at the time of discharge and CASE scores were noted. The optimal cut-off values for NLR, MLR, NPR, PLR, and SII were 4.87, 0.35, 0.032, 122.83, and 1566.69 respectively ( Table 2 ). Univariate analysis showed that compared with the good prognosis group, the poor prognosis group had higher elevated level of NLR, MLR, NPR, and SII at discharge. Multivariate logistic regression analysis showed older age at onset, dyskinesia/dystonia, gait instability and ataxia, higher level of NPR were independent risk factors for the poor prognosis of AE patients ( Table 4 ).