To the Editor:
According to the World Health Organization Global Tuberculosis Report from 2013, there were 8.6 million incident tuberculosis (TB) cases globally and India alone contributed 26% to this global scenario [1]. Of the five countries with the largest number of TB incident cases in 2012, India tops the list [1]. Epidemiological data suggest that extrapulmonary TB (EPTB) constitutes about 15–20% of all TB cases, but among HIV-TB co-infection it accounts for 50% of the cases [2]. Out of 1 183 373 new TB cases notified globally, 234 029 (20%) were reported to be cases of EPTB [1].
Difficulty in sampling from the extrapulmonary sites and the paucibacillary nature of the specimens make EPTB a diagnostic challenge. Dependency on smear microscopy in these samples may lead to higher false negative rates due to the low sensitivity of this technique. Mycobacterium tuberculosis (MTB) culture is quite a protracted technique, requiring well-trained laboratory personnel, and delay in diagnosis can cause more harm as the treatment is often started empirically.
Rapid nucleic acid amplification tests are emerging extensively to provide better yield for rapid diagnosis of TB. The Xpert MTB/RIF (Cepheid, Sunnyvale, CA, USA) is an automated, hemi-nested real-time PCR for detecting MTB complex and rifampin (RIF) resistance, which was initially evaluated for pulmonary specimens in large studies [3–5].
The present communication reports the performance of Xpert MTB/RIF in EPTB samples, with a large sample size from a single centre in a country with a high TB burden.
All EPTB samples from indoor as well as outdoor facilities of the All India Institute of Medical Sciences (AIIMS) hospital, New Delhi, India, were received in the Tuberculosis Laboratory (accredited Intermediate Reference Laboratory for Delhi National Capital Region by the Ministry of Health and Family Welfare, Government of India) of the Dept of Internal Medicine, AIIMS, New Delhi. The AIIMS Institutional Ethics Committee approved the study (IEC/NP-441/2012&RP-08/2012). All adult subjects with clinical suspicion of EPTB were enrolled and were either treatment naïve or were on anti-TB treatment for ≤2 weeks. A total of 1376 samples from 1274 patients were included in the study and written informed consent was taken from study subjects. The samples were subjected to Ziehl Neelsen staining, Xpert MTB/RIF assay and culture inoculation on both BACTEC Mycobacteria Growth Indicator Tube (Becton Dickinson, Sparks, MD, USA) for liquid and Löwenstein–Jensen media for solid culture. Drug susceptibility testing (DST) was carried out on Löwenstein–Jensen media. After exclusion of contaminated cultures (n=67), insufficient samples (n=8) and GeneXpert “error” results (n=9), 1292 samples were analysed for the study. The laboratory staff performing the Xpert MTB/RIF test was unaware of the solid culture DST results and vice versa.
Data were analysed using STATA statistical software version 12.1 (StataCorp LP, College Station, TX, USA). Sensitivity, specificity and negative and positive predictive values were calculated. The reference standards for each sample type were both culture and a composite reference standard (CRS). The CRS included parameters like smear, culture, histology and cytology reports (for biopsy samples and aspirates, respectively), biochemical tests such as adenosine deaminase levels (for pleural fluid, ascitic fluid, pericardial fluid and cerebrospinal fluid (CSF)) and response to treatment during follow-up visits.
The overall sensitivity and specificity of Xpert MTB/RIF assay with culture were 71% and 95%, respectively. With CRS, the sensitivity increased from 50% to 91% with increasing conventional diagnostic parameters (of the CRS) taken as positive, which clearly suggests that if the result by Xpert MTB/RIF test is positive for a sample, the case is more likely to be a true case of TB (table 1). These sensitivities and specificities, when translated into likelihood ratios (LR), accounted for LR+ of 13.7 and LR- as 0.30 with culture.
A positive Xpert MTB/RIF result also indicates RIF-sensitive or -resistant MTB, subsequently compared with the phenotypic DST results. Out of the 241 Xpert and culture-positive samples, 27 samples were RIF resistant by DST, out of which 26 were resistant but one was sensitive for RIF by Xpert MTB/RIF. 211 samples were sensitive from both Xpert MTB/RIF and phenotypic DST, three were sensitive by phenotypic DST but RIF resistant by Xpert. This accounted for a total sensitivity of 96.3% (26 out of 27) and specificity of 98.6% (211 out of 214).
Vadwai et al. [6] demonstrated a sensitivity of 83% and specificity of 73% for 533 EPTB patients. Their study had a higher proportion of biopsy and cold abscess samples than various body fluids, which might account for the higher sensitivity in their study. The reason for difference in the specificity of the present study and the study by Vadwai et al. [6] could be attributed to the inclusion criteria of patients and the sample size. In the present study, we excluded patients who were already on anti-TB treatment for >2 weeks. The overall performance of Xpert MTB/RIF in our study might vary from different studies across the globe because of the variation in proportion of sample types [7–15].
In this study, it was observed that the Xpert assay detected 71% of the “confirmed TB” cases where culture and response to anti-TB treatment were positive. It also identified 68% of “possible TB” cases where culture, biochemical and histopathology reports were negative and only the response to anti-TB treatment was positive. Of the cases where all parameters were negative, Xpert MTB/RIF detected 0.8% of these cases as positive. High specificity of the assay in all the specimens explains the low false positivity achieved by this diagnostic tool, which can thus be a useful rule-in test for EPTB diagnosis.
Suboptimal performance of the assay was observed for various body fluids, with sensitivity ranging from 18% to 40%, which is consistent with several previous studies [7–12]. The present study also indicates exemplary performance of the assay in patients with cold abscesses and lymph nodes and these findings are similar to previously published literature [7–9, 13]. The difference in sensitivity could be attributed to the fact that in lymph nodes and its aspirates the bacteria are localised to the site of infection, whereas in body fluids, presence of PCR inhibitors and the paucibacillary nature of the specimens may result in lower sensitivity.
TB meningitis poses a diagnostic challenge to clinicians, as it results in high morbidity and mortality with consequent sequelae. This is due to the lack of optimal rapid diagnostic techniques. A reasonable sensitivity of 68% was observed in CSF by Xpert MTB/RIF assay and a few published studies have also reported similar findings [14, 15], suggesting that the assay fares well for diagnosing TB meningitis cases.
There are several limitations of the Xpert MTB/RIF assay. It only detects RIF resistance or sensitivity and the former is a surrogate marker for multidrug-resistant (MDR) TB. This might lead to over-estimation of MDR-TB even in regions where RIF mono-resistance is high. In this scenario, line probe assay has an edge over Xpert MTB/RIF assay. Xpert is also sensitive to high temperature and humid conditions, which are quite prevalent in countries with a high TB burden, like India. Despite these limitations, introduction of this molecular DST technique in the current EPTB diagnostic algorithm will not only help in early diagnosis of TB but will also provide information about MDR-TB.
To conclude, the performance of Xpert MTB/RIF varies with the sample type but it is a promising diagnostic test for lymph nodes, cold abscesses and CSF specimens. In case of serosal fluids, which constitute a major challenge in EPTB, the diagnostic utility of the assay is limited.
Acknowledgments
The authors acknowledge the help of Peter Small (Deputy Director of the Tuberculosis Program, Bill and Melinda Gates Foundation, Seattle, WA, USA), Puneet Dewan (Senior Program Officer, TB, Bill and Melinda Gates Foundation, Delhi, India) and K.S. Sachdeva (Additional Deputy Director General, Central TB Division, Ministry of Health and Family Welfare, Government of India) for their valuable inputs in the study design. We are thankful to Vinod Raikwar and Sukhbir Singh (both Dept of Internal Medicine, All India Institute of Medical Sciences (AIIMS), New Delhi, India) for sample collection and sample processing in the laboratory. The authors are also thankful to the AIIMS administration for providing infrastructure, financial aid, lab space and other logistical help. We would also like to thank the faculty and residents from the AIIMS Depts of Internal Medicine, Neurology, Gastroenterology, Obstetrics and Gynaecology, Surgical Disciplines and ENT for sending the samples and contributing to the study.
Footnotes
Support statement: Surendra K. Sharma was supported by a J.C. Bose National Fellowship (SB/S2/JCB-04/2013), from the Ministry of Science and Technology, Government of India.
Conflict of interest: None declared.
- Received February 19, 2014.
- Accepted June 19, 2014.
- ©ERS 2014