Validation of a diagnostic algorithm for adult tuberculous meningitis

Area of Science:

• Neurology
• Infectious Diseases
• Medical Diagnostics

Background:

• Tuberculous meningitis (TBM) presents diagnostic challenges, particularly in differentiating it from other forms of meningitis.
• Low cerebrospinal fluid (CSF) glucose is a common indicator in patients with meningitis, necessitating accurate diagnostic tools.

Purpose of the Study:

• To prospectively evaluate a diagnostic algorithm for tuberculous meningitis (TBM) in HIV-negative patients with low CSF glucose.
• To compare the performance of logistic regression method (LRM) and classification and regression tree (CART) in diagnosing TBM versus bacterial meningitis (BM).

Main Methods:

• Prospective evaluation of 205 HIV-negative patients presenting with meningitis and low CSF glucose.
• Classification of patients into TBM or BM groups using logistic regression method (LRM) and classification and regression tree (CART).
• Comparative analysis of diagnostic performance for TBM vs. BM and TBM vs. non-TBM, including microbiologically confirmed cases.

Main Results:

• The diagnostic algorithm achieved high sensitivities for TBM: 99% with LRM and 87% with CART.
• Sensitivities for bacterial meningitis (BM) were also substantial: 81.5% (LRM) and 86.5% (CART) in the primary analysis.
• Similar diagnostic rates were observed in microbiologically confirmed cases, indicating robust performance.

Conclusions:

• The evaluated diagnostic algorithm demonstrates superior performance compared to microbiological confirmation rates in routine laboratories.
• The algorithm is recommended for use in high-prevalence TB settings with limited diagnostic facilities.
• Further validation in HIV-endemic settings is necessary to confirm its utility across diverse populations.