Development of a prediction model integrating PD-1 and ICOS for the early differential diagnosis between autoimmune and viral encephalitis

Kaiyue Xu ¹ Jinjing Jia ¹ Yinghui Duan ¹ Shuying Chen ¹ Xinyi Xiao ¹ Feng Zhu ¹ Xin Wang ¹ Yanzheng Gu ² Jingluan Tian ^1* Qun Xue ^1,2*

• ¹ Department of Neurology, The First Affiliated Hospital of Soochow University, Suzhou, China
• ² Jiangsu Institute of Clinical Immunology, First Affiliated Hospital of Soochow University, Suzhou, Jiangsu Province, China

AbstractBackground: Early diagnosis and treatment for encephalitis are crucial for improving patient outcomes and reducing the economic burden, especially given the overlapping symptoms and low specificity of auxiliary diagnostic tests between viral encephalitis (VE) and autoimmune encephalitis (AE). Since these two conditions require different treatment approaches, an early differential diagnosis between AE and VE is a critical challenge. Methods: This study enrolled a cohort of 75 patients (38 with VE and 37 with AE) be-tween September 2022 and July 2024. The demographic data, clinical characteristics and laboratory test results were collected. The expression levels of co-stimulatory mole-cules were detected by flow cytometry and ELISA within 7 days for viral enceph-alitis and 90 days for autoimmune encephalitis in the early phase of the disease. Dif-ferential analysis, logistic regression analysis, and Lasso regression were employed for model construction. Finally, a nomogram and a Receiver Operating Characteristic (ROC) curve were developed to visualize the model and evaluate its predictive accura-cy. Results: Upon analyzing the collected data, a model for the early differential diagnosis between AE and VE was eventually established. This comprehensive model incorpo-rates ten variables: serum creatinine and chloride levels, the percentage of peripheral blood CD4+ICOS+ and CD19+PD-L1+, plasma sICOSL (soluble Inducible Costimula-tory Ligand), cerebrospinal fluid (CSF) glucose content, and the presence of fever, nau-sea, vomiting, as well as headaches and cognitive impairment. Among them, patients with creatinine <60.75 (μmol/L), chloride <106.25 (mmol/L), CD4+ICOS+ ≥11.2%, CD19+PD-L1+ ≥12.35%, plasma sICOSL≥286.37 ng/mL, CSF sugar content ≥3.775 (mmol/L), and those with cognitive impairment are more likely to be diagnosed with AE. The AUC-ROC of our model is 0.942 (95% CI: 0.887-0.997), with a sensitivity of 0.844 and a specificity of 0.971, indicating strong diagnostic performance.Conclusion: This diagnostic model offers a convenient tool for distinguishing AE from VE in the early phase, facilitating early diagnosis and treatment, improving patient prognosis, and reducing financial burdens.