Background
Early diagnosis and treatment of pulmonary tuberculosis is essential in reducing the spread, morbidity, mortality and the escalating costs associated with advanced disease [
1]. Microbiological confirmation of
Mycobacterium tuberculosis disease can be challenging in patients with a low bacterial load or in those who do not expectorate. Bronchoscopy with washings, with or without biopsy, can aid diagnosis by providing adequate samples for testing [
2,
3].
Most of the evidence confirming the utility of the newly WHO endorsed Xpert MTB/RIF has been based on studies performed on sputum samples [
4,
5]. The reported sensitivity of 90.4 % is significantly higher than that of smear microscopy, which is around 28 % [
4,
6]. Recent studies have also suggested that Xpert MTB/RIF’s sensitivity is not affected by HIV status [
4,
7]. There is limited data validating the routine use of Xpert MTB/RIF on bronchial lavages, as the majority of cohorts contained low numbers of patients from whom bronchial lavages were obtained [
4,
7‐
9].
The aim of this study was to evaluate the diagnostic efficiency of Xpert MTB/RIF performed on bronchial washing fluid obtained bronchoscopically from patients with a clinical and radiological suspicion of pulmonary tuberculosis.
Methods
Study population
We screened all smear negative/sputum scarce patients who underwent bronchoscopy as part of their work up for suspected pulmonary tuberculosis at Tygerberg Academic Hospital and Kuils River Hospital during the period from January 2013 to April 2014, and retrospectively included all patients with complete sets of medical, radiological and laboratory data. Tygerberg Academic Hospital is a tertiary hospital in Cape Town, South Africa. It is one of two referral centres and it renders a tertiary service to a population of approximately 3.6 million people. Kuilsrivier Hospital is a secondary hospital in Cape Town, offering private health care to patients. For the purposes of our study we considered a clinical suspicion of pulmonary tuberculosis if any two of the following were present: known HIV infection, persistent cough lasting > 3 weeks, haemoptysis, weight loss > 4 kg, intermittent fever > 3 weeks or drenching night sweats > 2 weeks. In addition, at least one of the following radiological criteria had to be present for inclusion: cavitation, diffuse infiltrates, hilar or mediastinal adenopathy. The study was approved by the Health Research Ethics Committee of Stellenbosch University (S13/04/063). A waiver of consent was approved due to the retrospective nature of the study.
Bronchoscopy procedure
Flexible bronchoscopies were performed in dedicated suites under conscious sedation. Washings were obtained by instilling 10 ml of 0.9 % saline into a segment of an affected lobe and aspirating available fluid (repeated three times). Transbronchial biopsies and transbronchial needle aspiration of lymph nodes were performed at the discretion of the endoscopist.
Microbiology
Samples were processed according to standard laboratory protocols by decontamination with N-acetyl-L-cysteine/NaOH, centrifugation and fluorescent microscopy for acid fast bacilli [
10]. Decontaminated samples were inoculated in the BACTEC MGIT culture system, (Becton Dickinson, Sparks, Maryland, USA) and incubated for 6 weeks [
11]. Positive cultures were identified as
M. tuberculosis and tested for susceptibility to rifampicin and isoniazid using the MTBDRplus line probe assay (Hain LifeSciences, Nehren, Germany).
All Xpert MTB/RIF samples were processed according to the manufacturer’s specifications [
12]. Bronchial washing fluid was incubated, without prior decontamination, with sample reagent in a cartridge and loaded within 30 min of preparation. A test was positive if
Mycobacterium tuberculosis was identified within 38 cycles [
13]. Xpert MTB/RIF generated results (cycle threshold values) were categorised as either high positive (< 16), medium (16–21), low (22–28) or very low positive (> 28) [
12].
Diagnosis of pulmonary tuberculosis
We use a positive culture of Mycobacterium tuberculosis as the gold standard for the diagnosis of pulmonary tuberculosis in our primary analysis. In a secondary analysis, we also included cases with histological or cytological features of caseating granuloma containing acid fast bacilli, or cases where at least three clinical and one radiological criterion were present in a patient who subsequently showed documented clinical and radiological improvement after six months of anti-tuberculosis treatment (despite negative microbiology).
Statistical analysis
Descriptive statistics were performed and a composite reference standard used. Analyses were conducted of data collected per patient. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of Xpert MTB/RIF were calculated using standard methods and Xpert MTB/RIF generated results (cycle threshold values) listed. Unless stated otherwise, data is displayed as means and standard deviation (SD).
Discussion
In this study, performed in a population with a high pre-test probability for pulmonary tuberculosis, we found Xpert MTB/RIF to have a sensitivity, specificity and negative predictive value of around 90 %, and a positive predictive value of approximately 80 %. Xpert MTB/RIF had a very high diagnostic efficiency of 89.3 %. The majority of Xpert MTB/RIF positive, culture negative results were observed in cases with very low positive Xpert MTB/RIF results and a past history of pulmonary tuberculosis.
The sensitivity observed in the present study is on par with sensitivities recently reported in two studies (ranging from 82–93 %), but the specificity is marginally lower than the reported 96–100 % [
7,
8]. The studies reported by Lee
et al. and Theron
et al. included similar numbers of patients diagnosed with pulmonary tuberculosis, and utilised comparable diagnostic criteria for the diagnosis of pulmonary tuberculosis [
7,
8].
We identified 9 cases of Xpert MTB/RIF positive culture, negative cases, of which three were treated for pulmonary tuberculosis. Current evidence suggests that pulmonary tuberculosis treatment within the preceding 5 years can potentially RIF/RIF [
14,
15]. Up to 27 % of patients have been reported to remain sputum Xpert MTB/RIF positive 26 weeks after successful anti-tuberculous treatment was initiated [
16,
17]. Due to the nature of the polymerase chain reaction test, Xpert MTB/RIF amplifies any DNA whether it originates from alive or dead bacilli. Therefore it cannot be assumed, solely on the basis of the test, that a positive result equates to active disease [
12,
18]. It is also unclear whether the bacterial load in bronchoscopy samples in patients with latent infection or recent exposure would reach this limit and potentially result in a very low positive Xpert MTB/RIF test result. More evidence is needed on the interpretation of Xpert MTB/RIF in the setting of recently treated tuberculosis, which may be the most important cause of false positive results.
No cross-reactivity has been reported between Xpert MTB/RIF and multiple bacteria, viruses, fungi or mycobacteria species other than tuberculosis and should therefore not be the reason for false positive results [
19,
20]. We observed a single case of false positive Xpert MTB/RIF in a patient with confirmed
Mycobacterium avium intracellulare infection. It is likely that dual pathology (including undiagnosed previous tuberculosis) and laboratory contamination may have been present in this case, as well as the other cases not treated as pulmonary tuberculosis, a finding that has been reported in similar settings with a high HIV prevalence [
21-
23].
It should be noted that 7 of the 9 Xpert MTB/RIF positive culture negative cases had very low positive test results. Some correlation between Xpert MTB/RIF generated quantitative information and bacterial load and disease severity has been described [
21,
22]. Xpert MTB/RIF’s calculated analytical limit of detection has been reported as 131 cfu/ml (95% CI 106.2–176.4). The relationship between Xpert MTB/RIF generated cycle threshold values, bacterial load and smear microscopy grades has shown that very low positive test results (cycle threshold values >28 cycles) correlate with negative microscopy and low positive values (22–28) with scanty positive smear microscopy [
7,
21,
22]. This also correlated to the lower end of time-to-positivity of cultures. More research is required to aid in the interpretation of Xpert MTB/RIF with very low positive results, which in clinical practice, may be interpreted in the same light as a negative or, at best, scanty positive smear microscopy result. Our observations suggest that very low positive results require confirmation by culture or other means in cases with a low pre-test probability. Cut-off values for positivity should potentially vary between high and low endemic tuberculosis areas with lower cycle threshold values accepted as positive in high endemic areas. Moreover, clinicians practicing in high burden settings should be aware that past tuberculosis and dual pathology should be considered in cases with unexpected positive Xpert MTB/RIF.
The present study has several limitations. The retrospective nature limited the data collection to cases with complete clinical, radiological and microbiological data, and precluded any analyses on the excluded patient population. Some cases with false negative cultures may have gone undetected, and the true sensitivity may be different. We were also not able to document the true impact of Xpert MTB/RIF performed on bronchial lavages on the time to diagnosis or treatment, which is arguably of even greater clinical relevance.
Conclusion
Xpert MTB/RIF has a higher sensitivity than smear microscopy and similar specificity for the immediate confirmation of pulmonary tuberculosis on specimens obtained from bronchial washing, and should routinely be utilised in patients with a high clinical suspicion of pulmonary tuberculosis. Positive Xpert MTB/RIF, culture negative results should be interpreted in the clinical context, and ideally be confirmed with additional tests and/or follow-up.
Competing interests
The authors declare that they have no competing interests.
Authors contributions
DAB contributed to the study design, collection of data, analysis and interpretation of data, and writing the manuscript. EMI, JWB, ACW, JDD and DP contributed to the collection of data and editing the draft. CFNK contributed to the study design, study supervision, analysis and interpretation of data, and writing the manuscript. All authors read and approved the final manuscript.