Background: Confirming the clinical suspicion of tuberculous meningitis (TBM) has always been problematic. Whilst smear and culture positivity are diagnostic, these tests have low sensitivity. The polymerase chain reaction (PCR) assay has given variable results.
Aim: This study attempted to improve the diagnostic yield by: (a) increasing the cerebrospinal fluid (CSF) volumes; (b) testing the yield from three specimens of CSF assumed to represent lumbar, cervico-thoracic cord, and base of brain CSF samples; (c) undertaking PCR assays using multiple primer sets; and (d) using real-time PCR.
Method: Patients suspected of having cranial or spinal meningeal tuberculosis were entered into the study. Three aliquots of CSF were subjected to smear, culture, and conventional and real-time PCR. Three sets of primers - IS6110, MPB64, and PT8/9 - were used. Patients were retrospectively classified into four categories: 'definite TB' (culture positive), 'probable TB' (clinical and other tests suggestive of TB), 'not TB', and 'uncertain diagnosis'.
Results: A total of 68 patients were studied. There were 20 patients classified as definite TB, 24 probable TB, 17 not TB, and seven uncertain diagnosis. Forty-eight of 57 (84.2%) patients tested were HIV seropositive. The IS6110 PCR was positive in 27 patients which included 18/20 culture positive cases, six in the probable TB group, and three in the not TB group. The MPB64 and PT8/9 primers did not increase the yield. Real-time PCR was positive in seven additional patients. Combining the definite and probable TB, the sensitivity of all PCR assays was 70.5% (31/44) and specificity 87.5% (21/24).
Conclusion: Targeting multiple sites of the TB genome using conventional PCR did not increase the number of positive cases. Real-time PCR was more sensitive. However, all the current techniques are still too insensitive to confidently exclude the diagnosis on laboratory grounds.