Abstract:

Objective: To explore the apllication of several cerebrospinal fluid (CSF) indicators to improve the detection sensitivity and achieve rapid screening and early treatment control of tuberculous meningitis (TBM). Methods: From 2011 to 2019, according to the inclusion and exclusion criteria, 100 TBM patients were randomly selected from 383 suspected TBM patients admitted to tuberculosis designated medical institutions in Suzhou of Jiangsu Province, Zigong of Sichuan Province and Guiyang of Guizhou Province. And 100 non-TBM patients were selected based on age and sex matched of 1∶1. The biochemical indicators such as white blood cell count, lactate, total protein, adenosine deaminase, and glucose in CSF were collected. Ultra high-throughput liquid chromatography was used to separate the metabolites of CSF. The metabolic indicators with statistically significant differences in distribution were screened by Orthogonal Projections to Latent Structures-Discriminant Analysis (OPLS-DA). A classification and regression tree (boosted-CART) model was constructed by machine learning. The effect of characteristic indicators on rapid screening of TBM patients with various types was evaluated. Results: Using the criteria of P<0.05, OPLS-DA variable important in projection value >1 and absolute value of Forld Change >1, the above model was combined to screen CSF L-glutamine and glucose levels as characteristic indicators. The cutoff values were 607.06 mmol/L (L-glutamine) and 60.03 mg/dl (glucose), respectively. The combination of the above two indicators improved the sensitivity of detecting probable and possible TBM from 66.7% (95%CI: 24.1%-94.0%) to 83.3% (95%CI: 36.5%-99.1%),AUC increased from 0.667 (95%CI: 0.444-0.889) to 0.708 (95%CI: 0.494-0.923)(Z=0.261, P>0.05). Conclusion: Biochemical and metabolic indicators such as CSF L-glutamine and glucose can improve the sensitivity of rapid TBM screening, which is conducive to the early diagnosis and treatment of TBM and to control the spread of tuberculosis.

Key words: Tuberculosis, meningeal, Cerebrospinal fluid, Brain diseases, metabolic, Early diagnosis, Communicable disease control