We evaluated microbiological diagnosis of tuberculous (TB) meningitis in a referral hospital in Indonesia. Over a ten-year period, we examined cerebrospinal fluid (CSF) samples of 1180 adult meningitis suspects. Sensitivity of different methods was compared, and results were stratified for HIV status, disease severity, and CSF volume. TB meningitis was bacteriologically confirmed in 501 patients. Using clinical diagnosis as reference standard (n = 713), sensitivity of different methods was 12.2% (86/703) for microscopy, 42% (73/174) for Xpert MTB/RIF, 46.0% (163/354) for solid culture, 48.8% (332/680) for liquid culture, and 64.0% (212/331) for in-house PCR. Head to head comparisons in 654 patients showed a higher yield of in-house PCR (32.3%) compared to culture (15.6%, P < 0.01). Microscopic observation of drug susceptibility (MODS) culture more rapidly became positive compared to other culture methods. Yield of culture was lower in HIV-infected (39/105) than in HIV-negative patients (N = 316/585; P < 0.01). Molecular and culture methods gave higher yields in patients with more severe disease (P < 0.01). CSF volume of ≥6 ml increased the yield of culture (42.8% versus 12.1% for CSF <6 ml, P < 0.01) and ZN-microscopy (18.3% versus 1.9% for CSF <6 ml, P < 0.01). CSF centrifugation had no clear effect on sensitivity of Xpert MTB/RIF. ZN-microscopy lacks sensitivity for diagnosis of TB meningitis. For molecular assays, in-house IS6110-PCR is more sensitive than Xpert MTB/RIF. MODS culture has a clear advantage in terms of speed. Large CSF volumes are necessary for all tests. The effect of CSF processing for Xpert MTB/RIF needs further study.
Keywords: Adult; Diagnosis; Meningeal; Microbiology; Tuberculosis.
Copyright © 2018. Published by Elsevier Inc.