Immune-inflammatory profiles and disease severity in pulmonary tuberculosis complicated by central nervous system tuberculosis

Objective To compare peripheral blood immune–inflammatory profiles between patients with pulmonary tuberculosis complicated by central nervous system tuberculosis (PTB + CNS-TB) and those with pulmonary tuberculosis alone (PTB), and to identify laboratory parameters associated with disease severity in PTB + CNS-TB. Methods This retrospective study included 4,829 inpatients with tuberculosis (January 2018–March 2026), comprising a PTB group ( n = 4,578) and a PTB + CNS-TB group ( n = 251). Propensity score matching (1:4) was performed using sex, age, body mass index, albumin, and hemoglobin, yielding 246 PTB + CNS-TB patients and 954 matched PTB controls. PTB + CNS-TB cases were further stratified by Medical Research Council (MRC) grading (Grade I: n = 165; Grades II–III: n = 86). Covariate-adjusted analyses with FDR correction were used to screen severity-associated parameters; LASSO and Firth’s logistic regression were applied to identify mutually independent factors and to construct a severity assessment model. Results After matching, PTB + CNS-TB patients exhibited a distinct bidirectional pattern characterized by reduced peripheral lymphocyte counts and lower peripheral T-cell counts (CD3 + , CD4 + , and CD8 + ) and CD45 + lymphocytes counts, alongside increased neutrophil-related inflammatory indices (including neutrophil count, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and systemic immune-inflammation index) (all P < 0.05). Within the PTB + CNS-TB cohort, severe disease was associated with higher white blood cell and neutrophil-related measures (notably NEU% and NLR) and more pronounced cerebrospinal fluid (CSF) abnormalities, particularly elevated CSF protein and decreased CSF chloride, whereas peripheral T-cell subset parameters did not differ significantly by severity. Multivariable modeling identified CSF protein, NEU%, and albumin as mutually independent factors associated with severe disease. A combined model including these three markers showed good internal discrimination (AUC = 0.772; bootstrap-corrected AUC = 0.767) and calibration. Conclusions PTB + CNS-TB is associated with reduced peripheral lymphocyte/T-cell subset counts accompanied by heightened a neutrophil-predominant inflammatory profile. Among PTB + CNS-TB patients, disease severity shows a closer association with neutrophil-related indices and routine CSF parameters, particularly CSF protein, than with peripheral T-cell subset counts. A model incorporating NEU%, CSF protein, and albumin may aid in severity assessment.