EBUS-TBNA in the rapid microbiological diagnosis of drug-resistant mediastinal tuberculous lymphadenopathy

Discussion

Mediastinal lymphadenitis is a common finding in pulmonary TB and may by itself account for up to 40% of extrapulmonary TB cases [7]. The diagnosis of intrathoracic TBLA traditionally relied on clinical and radiographic findings, as well as a therapeutic trial with anti-tubercular drugs. However, this can lead to redundant treatment, risk of adverse effects and a delay in the actual diagnosis and treatment. Microbiological and histopathological examination of the lymph nodes is indispensable for obtaining a definitive diagnosis, making sampling of the lymph nodes essential [7].

Mediastinoscopy was long considered the gold standard for sampling mediastinal lymph nodes, but it is invasive and requires general anaesthesia. Also, posterior subcarinal and hilar lymph nodes, which are commonly affected in TB, cannot be accessed. EBUS-TBNA enables sampling of the mediastinal lymph nodes under real-time ultrasonographic guidance, making it safe and minimally invasive, and facilitating adequate sampling [20]. It compares well with conventional TBNA and mediastinoscopy, with a sensitivity of 80% and specificity of 100% for the detection of TB by both microbiological and histopathological examinations [21].

The rate of positive cultures from EBUS-TBNA specimens is 20%–63%. A recent meta-analysis reported a culture positivity rate of 54% and a smear positivity rate of 30% [22]. Another meta-analysis showed that the pooled sensitivity and specificity of EBUS-TBNA for the diagnosis of intrathoracic TB were 0.80 and 1.00 respectively [21]. In the current study, 25 patients were Xpert MTB/RIF positive but TB culture negative. Several previous studies have identified specimens that were Xpert MTB/RIF positive and TB culture negative [23, 24]. High Xpert MTB/RIF positivity and TB culture negativity has been attributed to low viable counts or non-viable tubercle bacilli due to either the decontamination procedure or the nature of the caseous lesion in the lymph node tissue, which may have contained dead tubercle bacilli [25]. The low sensitivity of culture as compared to that of Xpert MTB/RIF has also been attributed to the paucibacillary nature of extrapulmonary specimens with uneven distribution of bacilli and formation of clumps [26]. Such cases (Xpert MTB/RIF positive and TB culture negative) are likely to be true positives as corroborated by the high specificity of Xpert MTB/RIF [23, 25, 27]. In the current subgroup of Xpert MTB/RIF-positive and TB culture-negative patients, 22 of 25 patients showed granulomatous inflammation and three patients showed chronic inflammatory changes on histopathology. This suggests that there was a high correlation of Xpert MTB/RIF positivity with histopathological findings. Because the current study setting is endemic for TB, patients were treated for TB based on Xpert MTB/RIF results, histopathological findings and clinical symptoms, despite having a negative TB culture. Similar approaches have previously been used with successful treatment outcomes [27]. The possibility of cross-contamination is least likely because the Xpert MTB/RIF assay is a closed system and is processed in accredited laboratories [23].

The Xpert MTB/RIF assay has an overall sensitivity of 88% and specificity of 99% for the detection of M. tuberculosis, and an overall sensitivity of 95% and specificity of 98% for the detection of rifampicin resistance [28]. The assay has been validated for extrapulmonary specimens in a high-prevalence endemic region like India, showing a sensitivity and specificity of 90.68% and 99.62% for the detection of M. tuberculosis, and sensitivity and specificity of 81.82% and 100% for the detection of rifampicin resistance [8]. The relatively low diagnostic yields of drug-resistant TB from culture and Xpert MTB/RIF in our study could be attributed to the paucibacillary nature of TBLA and the association between bacillary load in tissue and detection of rifampicin resistance [8]. Xpert MTB/RIF-negative and TB culture-positive samples were observed in 23% of patients. It has been postulated that the concentration of mycobacteria in the GeneXpert-negative samples is lower than the limit of detection of the Xpert MTB/RIF assay, resulting in the Xpert MTB/RIF-negative but TB culture-positive results [23].

We encountered patients in whom culture-based DST, but not Xpert MTB/RIF, revealed rifampicin resistance (figure 1). The Xpert MTB/RIF assay uses five molecular probes complementary to the wild type rpoB gene in the rifampicin-resistance-determining region (RRDR) of the M. tuberculosis genome. Rifampicin-resistance mutations that inhibit binding of these probes can be detected in up to 95% of cases [8]. It is therefore possible to miss drug-resistance mutations in the remaining 5%, which may either lie in the RRDR and only delay binding of the probes, or lie outside the RRDR altogether [29].

Dhasmana et al. [30] reported three cases of MDR-TB out of 88 cases with microbiologically confirmed mediastinal TBLA: two were identified by Xpert MTB/RIF and one by line probe assay. Thangakunam et al. [31] did not find any rifampicin resistance on Xpert MTB/RIF testing of EBUS-TBNA aspirates. In comparison, we found that 20% of patients (11 out of 56) with microbiologically proven TB in our cohort had drug-resistant TB. This observation can be attributed to various factors, e.g. the high prevalence of MDR-TB in the region where the current study was undertaken or referral bias [32]. The current study does highlight the value of EBUS-TBNA specimens in the diagnosis of TB as well as drug-resistant TB. However, the actual incidence of drug-resistant TB in our TB-endemic region in EBUS-TBNA specimens would need to be assessed prospectively in a well-defined patient cohort.

Aspirate samples obtained from lymph nodes that had necrotising granulomas yielded a microbiological diagnosis. Sun et al. [33] reported that necrosis was a risk factor associated with a better chance of culture positivity. The tubercle bacilli reside in the necrotic regions of necrotising granulomas and sampling of these regions is important to obtain a diagnosis of TB.

The median days to diagnosis of drug-resistant TB by Xpert MTB/RIF was 1 day (range 1–3 days), and by DST was 30 days (range 21–60 days, including DST). The delay of a day or more for Xpert MTB/RIF testing in some cases was because the specimens were sent to an offsite laboratory. However, our findings are comparable to the findings of Navani et al. [13] (median time to positive culture, excluding DST, 16 days, range 3–84 days; one centre in the study extended incubation beyond 6 weeks) and of Dhasmana et al. [30] (mean time to positive culture for microscopy positive cases, excluding DST, 13.3 days, range 6–19 days).

The limitations of the current study include, first, the lack of a single adequate gold standard against which either Xpert MTB/RIF or MGIT 960-based DST may be measured. Positive results from either Xpert MTB/RIF or MGIT 960 were used as a reference standard and, indeed, taken as an indication to initiate MDR-TB treatment regimens in our patients. Supporting this was the evidence from literature documenting the high specificity and positive predictive values of both tests [15, 33]. Second, the study institute serves as a referral centre and several patients were transferred back to their referring physicians following EBUS-TBNA and consequently were not followed up. Therefore, the current study cohort is unable to provide the outcome information for these patients. Further analysis on a larger sample of patients with drug-resistant TB with follow-up of treatment outcome would be useful.

In conclusion, the current study highlights that EBUS-TBNA is a valuable tool in the diagnosis of mediastinal TBLA, including drug-resistant M. tuberculosis. The Xpert MTB/RIF assay carried out on EBUS-TBNA specimens provides a rapid diagnosis in a significant proportion of patients with TB. However, all EBUS-TBNA specimens should also be sent for culture-based DST, which provides a likely paucibacillary population missed by the molecular technique but also detects resistance to drugs other than rifampicin and rifampicin resistance that may be missed by Xpert MTB/RIF testing. Combined Xpert MTB/RIF and TB culture testing on EBUS-TBNA specimens is therefore a useful modality in the diagnosis of drug-resistant mediastinal TBLA.