In this study, clinical and widely available laboratory features were collected from 91 patients with a chronic meningitis syndrome to find out which of these were associated with a diagnosis of TBM. Based on these test results (positive for TB culture or AFB smear or basal meningeal enhancement on contrast CT scan of the head), only 44 would fulfil the criteria of definite TBM. If only growth on TB culture or positive AFB smear were used as the gold standard, then this number would be even lower (4/91). The final discharge diagnosis was TBM in 68 of the 91 cases, whereas there were other alternative diagnoses in the other cases. Other CNS infections were among the next most common final diagnoses: bacterial and viral meningitis (6·6% each) and cryptococcal meningitis (2·2%). Non-infectious causes (6/91) were also among the differential diagnoses, such as metabolic encephalopathy, hypoxia, and subarachnoid hemorrhage, but these are usually easily ruled out on the basis of history, PE, and simple laboratory examinations.
Due to the low yield of conventional TB culture or AFB smear in our setting, we chose to include basal meningeal enhancement on CT contrast study of the head as part of the gold standard criteria for TBM. Several authors have determined the validity of this approach: Kumar (1996) compared the CT findings of 94 children with TBM with those of 52 children with pyogenic meningitis, and concluded that basal meningeal enhancement, tuberculoma, or both, were 89% sensitive and 100% specific for the diagnosis of TBM. Przybojewski [5] showed that the criteria most clinically useful were the Y-sign, linear enhancement, contrast filling the cisterns, and asymmetry, due to their high likelihood ratios.
Due to the shortcomings of the laboratory examinations used as the gold or reference standard in this study, an alternate reference standard could have been used, such as a consensus panel diagnosis, wherein at least two experts would examine each chart and arrive at a consensus as to whether each patient did indeed have TBM. Important data such as the response to anti-Koch's medications could have been taken into account for the panel to arrive at a diagnostic decision with a confident level of certainty. However the panel would have to be blinded to the results of the tests being studied as possible predictive factors, lest incorporation bias come about and artificially inflate the odds ratios.
It is unfortunate that only abnormal CSF was found to be associated with the diagnosis of TBM, in contrast to previous studies. Thwaites, Kumar, Sunbul, Youssef, and Moghtaderi [1, 6–9], cited in the review of literature. This may be because of differences in the study population. In the earlier studies, the population was of patients presenting as a meningitis syndrome, whether acute or chronic; cases of TBM were compared with other bacterial meningitides, but arguably this would not be useful in an actual clinical setting because it is not often that clinicians have to discriminate between TBM and acute bacterial meningitis. More often the dilemma is to distinguish between TBM and other chronic meningitides such as cryptococcal meningitis. Also, the population in the present study would be more relevant for local clinicians, as the clinical presentation of TBM in the country may be different from places such as Vietnam and India, places with high HIV prevalence, where the previous studies were carried out. Coinfection with HIV increases the risk of TBM and modifies the clinical presentation: usually patients would present with acellular CSF and high AFB loads [6].
This study shows that diagnostic rules and prediction models which were developed in one setting may not necessarily be relevant in another setting, as the factors which were found relevant in previous studies turned out not to be significant in the local setting, most probably due to differences in the study population.
This study also shows that an abnormal CSF profile (lymphocytic pleocytosis + decreased glucose + increased protein) has a moderate sensitivity and specificity, in fact comparable with more advanced molecular methods such as PCR, and a definite improvement over the current gold standard of AFB smear or TB culture. Hence in the setting of a chronic meningitis syndrome, clinicians confronted with this finding may be reasonably confident of the diagnosis of TBM, and use the associated likelihood ratios to arrive at a decision as to whether to order more tests or proceed with anti-Koch's treatment. Also, clinicians confronted with these findings should first rule out those conditions which have entirely different therapies, such as cryptococcal meningitis and partially-treated bacterial meningitis, by doing the appropriate laboratory exams. In those patients where it is not immediately evident what the etiologic organism is, then the best course would be to treat for TBM and start anti-Koch's medications by default, as this is the most prevalent etiology of a chronic meningitis in the local setting, and then to also institute additional treatment for those other conditions which the clinician is ruling in. He may then modify his subsequent management based on the response to therapy, and a repeat of the pertinent laboratory exams where appropriate.
There are some limitations to this study:
1) The desired sample size was not achieved within the time frame of this study; conceivably some factors would turn out to be positively predictive if a larger population were studied;
2) There is subjectivity involved in assessing for the presence of basal meningeal enhancement on CT head contrast study; also inter-rater and intra-rater variability was not assessed. This would have affected the performance of the gold standard test.
3) Related to this, the yield of the conventional gold standard tests, which are AFB smear and TB culture, is still very low. If these were the only tests considered as part of the composite gold standard, then it would have been useless to proceed with the logistic regression. Some techniques that may be used to improve the yield of AFB staining include staining the clot that forms in standing CSF and spinning down the CSF sediment onto a slide for microscopic examination. In 1979, Kennedy and Fallon [10] showed that staining multiple samples of CSF enhanced the sensitivity to 86%.
4) Finally, the presence of HIV co-infection was not tested. As mentioned earlier, a high rate of co-infection with HIV may affect the clinical presentation and affect the choice of clinical and laboratory features to be studied.
