Rapid and accurate diagnosis of TBM are key factors for improving prognosis, but they are still challenging. The findings in this study showed good diagnostic performance for detecting M. tuberculosis by Xpert MTB/RIF assay with high sensitivity (83.33%), specificity (96.23%), PPV (83.33%), and NPV (96.23%) when using an MGIT TB culture as a reference standard for definite diagnosis. There were 3 cases in which M. simiae was isolated from the CSF culture with negative results from the Xpert MTB/RIF assay, which could emphasize the diagnostic accuracy. In comparison with a conventional AFB staining smear, despite using centrifuged CSF and highly skilled and experienced technicians for examination, our study found that the Xpert MTB/RIF assay had a significantly higher sensitivity and an insignificantly lower specificity. This phenomenon can possibly be explained by the difference in CSF volume for each examination (800 µL for the Xpert MTB/RIF assay versus 100 µL for smear)and the very low CSF volume for AFB staining, which might cause negative results from all specimens in this study20.
From multiple previous studies, the differences in research methods and laboratory techniques are attributed to the different results9,15–20. By excluding CNS cryptococcosis, our study modified these factors to improve diagnostic testing values. This is one of the major causes of lymphocytic meningitis in this region, and its clinical characteristics mimic TBM in the study. Another modification was the use of a CSF volume of 800 µL compared to 200 µL in the previous one in Vietnam9, which could be a cause of the higher sensitivity seen in our study (83.3% Vs. 59.3%).
The TBM score had very high sensitivity but low specificity, causing high rates of false positives when using 6 points, the lowest score being “possible TBM” which could be classified as the cut-off level for TBM diagnosis. The Xpert MTB/RIF had higher costs and required skilled technicians which might affect the test’s accessibility, especially in a resource-limited setting28. In our study, we employed sequential testing to address these issues. We used the TBM score as the first-stage screening test to increase the inclusion of suspected cases, and performed the second-stage screening test with the Xpert MTB/RIF assay when the score was 6 points or higher. With these sequential combinations, there was an improvement in the diagnostic performance, especially the specificity, which could reduce false positive cases when using the score alone. Moreover, the screening score might be useful in the prevention of “overuse” of a high-cost test in a resource-limited setting, such as the Xpert MTB/RIF.
In cases that required the treatment of TBM, the Xpert MTB/RIF assay had a significant agreement of 90.77% with a kappa of 0.743, statistically equal to the reference standard MGIT culture. According to statements that earlier diagnosis and treatment are better, with a shorter turnaround time than cultures, the use of the Xpert MTB/RIF assay on the CSF specimen for a second-stage screening test when the TBM score is at least 6 points could be a helpful diagnostic tool for TBM supporting the clinical decision to initiate prompt treatment with anti-tuberculosis drugs. In the past, we recognized these integrations as very important factors in reducing morbidity and mortality rates from delayed management.. This was combined with the cost-effectiveness of the initial screening with the TBM scoring system before using an Xpert MTB/RIF assay.
No discordance of rifampicin resistance was found between an Xpert MTB/RIF assay and a culture with drug susceptibility testing, but there was a case of INH mono-resistance found in our study. The Xpert MTB/RIF assay was established on the basis that M. tuberculosis with rifampicin resistance may also have isoniazid resistance14,15. However, there have been about 7–10% of isoniazid mono-resistance cases in Thailand which is a little concerning10,29. Considering this, we should continue to use the Xpert MTB/RIF assay in conjunction with the reference standard culture and drug susceptibility testing to identify drug-resistant M. tuberculosis.. Surprisingly, there was no multidrug-resistant M. tuberculosis isolated from the CSF specimens in our study which differed from previous data in our hospital and in northern Thailand10. For this reason, we cannot evaluate the diagnostic accuracy in a MDR-TB case or in the M. tuberculosis strain, which had low levels of rifampicin resistance. The subgroup analysis revealed that the HIV-infected group's Xpert MTB/RIF assay had higher sensitivity and specificity, but there was no statistical significance when compared to the non-HIV group. These results may be because there was too low a population of HIV-infected people, which will have affected the statistical power for detecting the statistical difference in our study. These issues can inspire further study for applying this assay to a specific population, especially in cases involving HIV infection.
Our study still has limitations. First, the inclusion criteria were not very clearly defined, and many cases with abnormal neuromanifestations from non-infectious causes, such as malignancies with CNS involvement, might be included in the study. Second, unreliable data about the duration of clinical syndrome obtained from patients with altered consciousness or their relatives affected the TB score calculation. All of these had an influence on the actual diagnostic test value. Finally, this study was conducted before the introduction of Xpert MTB/RIF Ultra, causing a missed opportunity to compare the diagnostic accuracy between both assays. Further studies with larger study populations, using more strict inclusion criteria, and comparing the diagnostic performance with the Xpert MTB/RIF Ultra assay are needed to make it through these limitations.
In conclusion, this study showed that the Xpert MTB RIF assay had high sensitivity and specificity in detecting M. tuberculosis from CSF specimens. The diagnostic accuracy improved when combined with a TBM scoring system in a two-step screening process. This was very helpful in making a rapid, accurate diagnosis and initiating prompt treatment for patients with tuberculous meningitis. Further studies to compare with the Xpert MTB/RIF Ultra assay are remarkably required.