Related Experiment Video
Updated: Jun 20, 2025
Flow Cytometric Analysis of Lymphocyte Infiltration in Central Nervous System during Experimental Autoimmune Encephalomyelitis
Published on: November 17, 2020
Logistic regression modeling of cytokines for cerebrospinal fluid evaluation in primary central nervous system
Di Wang1, Yejia Wu2, Jingjing Ma3
1Department of Laboratory Medicine, Huashan Hospital, Fudan University, Shanghai 200040, China.
Background:
The diagnostic utility of cerebrospinal fluid (CSF) cytology encounters impediments stemming from variability in cell collection techniques and pathologists' morphological acumen, resulting in wide-ranging CSF positivity rates for primary central nervous system lymphomas (PCNSL). Such disparity impacts patient evaluation, treatment stratagem, and prognostication. Thus, this study endeavors to explore liquid biomarkers complementary to CSF cytology or immunophenotype analysis in the diagnosis of CSF involvement.
Methods:
398 newly diagnosed PCNSL patients were categorized into CSF involvement and non-involvement groups based on CSF cytology and immunophenotype analysis. Binary logistic regression analysis was performed on 338 patients to investigate factors predicting CSF involvement and to develop a joint prediction model. An additional cohort of 60 PCNSL patients was recruited for model validation. Statistical analyses included the Mann-Whitney U test for comparing various CSF parameters between two groups. ROC curve analyses were performed for each biomarker to identify PCNSL CSF involvement.
Results:
The cytokine IL-10 level in CSF has emerged as the most promising biomarker for CSF evaluation, boasting an ROC AUC of 0.922. C-TNFα and soluble C-IL2R demonstrate efficacy in quantifying tumor burden within the CSF. Logistic regression identified C-IL10lg (OR = 30.103, P < 0.001), C-TNC (OR = 1.126, P < 0.001), C-IL2Rlg (OR = 3.743, P = 0.029) as independent predictors for CSF involvement, contributing to a joint predictive model with an AUC of 0.935, sensitivity of 74.1 %, and specificity of 93.0 %. Validation of the model in an independent cohort confirmed its effectiveness, achieving an AUC of 0.9713.
Conclusions:
The identification of these feasible biomarkers and the development of an accurate prediction model may facilitate the precise evaluation of CSF status in PCNSL, offering significant advancements in patient management.
