We prospectively enrolled adult patients aged ≥16 years who were suspected of having TBM in Asan Medical Center, a 2,700-bed tertiary hospital in Seoul, Korea, a country with an intermediate TB-burden, from April 2008 through March 2014 [45]. Using this prospective cohort, we evaluated the utility of the uniform case definition for differentiating between definite TBM, VM and BM. Patients with TBM were categorized as definite TBM (acid-fast bacilli seen on cerebrospinal fluid (CSF) microscopy, positive culture of Mycobacterium tuberculosis in CSF, or positive polymerase chain reaction (PCR) of M. tuberculosis in CSF), probable TBM (diagnostic score ≥12), and possible TBM (diagnostic score 6 - 11) according to the uniform case definition [1]. A diagnosis of definite VM was established if viral pathogens were identified by positive PCR in the CSF. Definite BM was confirmed by positive culture of bacteria in CSF or blood. The study protocol was approved by the Institutional Review Board of Asan Medical Center.

During the study period, 93 adult patients were treated for TBM; 27 (29%) were classified as definite TBM, 25 (27%) as probable TBM, and 41 (44%) as possible TBM. We also identified 23 PCR-confirmed VM and 20 culture-confirmed BM (Table 1). One patient with probable TBM had HIV infection. Most of enrolled patients with TBM, VM and BM underwent cerebral imaging except 2 patients with BM. Among the 27 patients with definite TBM, 22 (81%) were scored as probable TBM and 5 (19%) as possible TBM. Among the 23 patients with definite VM, 13 (57%) were scored as unlikely TBM and 10 (43%) as possible TBM. Of the 10 patients scored as possible TBM, 9 received scores of ≥4 on the clinical criteria (maximum 6) and 8 had scores of ≥3 on the CSF criteria (maximum 4). There was a trend for meningitis caused by varicella-zoster virus to be scored as possible TBM more often than other VM (60% [6/10] vs. 31% [4/13]; P = 0.22). Out of the 20 patients diagnosed with definite BM, 13 (65%) were scored as unlikely TBM, 1 (5%) as probable TBM and 6 (30%) as possible TBM. Of the 7 patients scored as probable or possible TBM, 3 received scores of ≥4 on the clinical criteria (maximum 6) and 6 had scores of ≥3 on the CSF criteria (maximum 4). The diagnosis of the 1 patient with BM having probable TBM score was neurosyphilis. Patients with neurobrucellosis and scrub typhus were included in BM patients received possible TBM score.

These findings indicate that some cases of VM and BM could masquerade as TBM despite the usefulness of the uniform case definition criteria, especially in severe cases of VM and indolent cases of BM. Overall, ‘probable TBM’ criteria differentiated definite TBM from both VM and BM and with a sensitivity of 81% (22/27) and specificity of 98% (42/43), while the ‘possible TBM’ criteria identified definite TBM with a sensitivity of 100% (27/27) and specificity of 60% (26/43). In detail, specificities of ‘probable TBM’ criteria for differentiating TBM from VM and BM, respectively, were 100% (23/23) and 95% (19/20). Specificities of ‘possible TBM’ criteria for differentiating TBM from VM and BM, respectively, were 57% (13/23) and 65% (13/20).

As the findings of Solomons et al. [23], our data suggest that the uniform case definition may be useful for differentiating TBM from VM and BM, but there was substantial overlap in clinical and laboratory features among them. Since the definition criteria were developed for research purposes that facilitating standardization of study cases [16], the scoring system should be interpreted cautiously in diagnosing TBM [7]. To avoid fatalities caused by delayed treatment and morbidity due to unnecessary exposure to anti-TB drug, a rapid and accurate tool for diagnosing TBM is urgently required.