Background: Bacteriological confirmation of tuberculous meningitis (TBM) is problematic, and rarely guides initial clinical management. A uniform TBM case definition has been proposed for research purposes.
Methods: We prospectively enrolled patients aged 3 months to 13 years with meningitis confirmed using cerebrospinal fluid analysis at Tygerberg Hospital, Cape Town, South Africa. Criteria that differentiated TBM from other causes were explored and the accuracy of a probable TBM score assessed by comparing bacteriologically confirmed cases to 'non-TBM' controls.
Results: Of 139 meningitis patients, 79 were diagnosed with TBM (35 bacteriologically confirmed), 10 with bacterial meningitis and 50 with viral meningitis. Among those with bacteriologically confirmed TBM, 15 were Mycobacterium tuberculosis culture-positive and 20 were culture-negative but positive on GenoType(®) MTBDRplus or Xpert(®) MTB/RIF; 18 were positive on only a single commercial nucleic acid amplification test. A probable TBM score provided a sensitivity of 74% (95%CI 57-88) and a specificity of 97% (95%CI 86-99) compared to bacteriologically confirmed TBM.
Conclusion: A probable TBM score demonstrated excellent specificity compared to bacteriological confirmation. However, 26% of children with TBM would be missed due to the limited accuracy of the case definition. Further prospective testing of an algorithm-based approach to TBM is advisable before recommendation for general clinical practice.