Utility of GeneXpert MTB/RIF assay for the diagnosis of pulmonary and extra-pulmonary tuberculosis, A report from Egypt

Early diagnosis of tuberculosis continues to be a challenge for clinicians. The World Health Organization (WHO) guidelines recommend the application of GeneXpert MTB/RIF in extra-pulmonary tuberculosis (EPTB) diagnosis. This study aimed to test and compare the accuracy of the GeneXpert MTB/RIF assay to diagnose pulmonary tuberculosis (PTB) and EPTB, compared to bacterial culture and to composite reference standard (CRS). The GeneXpert assay diagnosed tuberculosis (TB) in 19.5 % of patients. With reference to bacterial culture, the sensitivity of this assay for detection of the pulmonary and extra-pulmonary specimens was perfect. For pulmonary specimens, on using CRS; the detected sensitivity and specificity of the GeneXpert assay were 78.3 % and 99.1 %, respectively. However, for extra-pulmonary specimens, the sensitivity and specificity of the GeneXpert assay were 37.1 % and 99 %, respectively. In the current study, the GeneXpert assay showed almost perfect agreement with the bacterial culture for TB diagnosis. The diagnostic accuracy of the GeneXpert assay was high in ruling in, but not in ruling out of EPTB.


Introduction
Globally, tuberculosis (TB) persists as a current and leading health concern with an estimated 10 million new cases in 2017, and only 67 % of the cases (6.7 million cases) have been diagnosed. Furthermore, an estimated 6.6 million rifampicin-resistant (RIF-R) cases occurred, but only 2.0 million (30 %) were identified (WHO. 2018). The WHO has classified Egypt as a middle/low-level country according to TB prevalence. The estimated TB annual prevalence is 11/100 000 cases with smear-positive (SP) active pulmonary tuberculosis (PTB), and 24/100 000 cases Novel Research in Microbiology Journal, 2021 with all types of TB, as reported by Moussa et al., (2016).
In clinical practice, early diagnosis of TB continues to be a challenge for clinicians, especially with extra-pulmonary TB (EPTB), childhood TB, and TB patients co-infected with HIV (WHO. 2016). Virtually, EPTB may affect every part of the body away from the lungs (Sharma and Mohan, 2004). According to a previous study conducted by Tortoli et al., (2012), EPTB has varied clinical manifestations, and atypical presentation, and therefore requires a high index of clinical suspicion as reported recently by Tag Eldin et al., (2019). Besides, the EPTB is difficult to diagnose due to the smaller number of bacteria in the specimens (paucibacillary nature); difficulty in obtaining specimens from deep-seated organs, and inability to get an extra specimen, as revealed by Bankar et al., (2018).
In low-income countries, conventional methods such as Ziehl-Neelsen (ZN) smear microscopy is a cheap and rapid method for the detection of acid-fast bacilli; however, it has poor sensitivity and poor PPV (positive predictive value) (Chen et al., 2012). Though culture is the gold standard for diagnosis of TB, it often takes weeks to have the results, which causes significant delay. Furthermore, the deficiency of diagnostic infrastructure, experienced staff and specialized laboratories interfere with proper patients care and outcomes, and exacerbate the dilemma of EPTB diagnosis. Thus, recent studies conduct by Bankar et al., (2018), Rasheed et al., (2019) highlighted that rapid and early detection of Mycobacterium tuberculosis (MTB), and the multidrug resistance/rifampicin-resistant (MDR/RIF-R) strains is an obligation. The WHO. (2013) has endorsed GeneXpert MTB/RIF (Xpert) assay (Cepheid, CA, USA) for the PTB diagnosis, as it is highly sensitive and specific for CP (culture positive) TB. Additionally, several studies conducted by Tortoli et al., (2012); Pang et al., (2017) have recommended that GeneXpert assay has a hopeful efficacy in detecting EPTB, which assists in following the WHO guidelines.
According to Helb et al., (2010), the GeneXpert assay is an automated closed-cartridge system; easy to use, bio-safe, requires minimal training, and its results are acquired within two hours. Boehme et al., (2010) added that the test detects MTB and rifampicin resistance simultaneously, which can thus be used as a representative marker for MDR-TB.
The objectives of this study were to test the accuracy of the GeneXpert assay to diagnose PTB and EPTB, compared to culture on Lowenstein-Jensen (LJ) medium and to a composite reference standard (CRS). This is in addition to detecting the prevalence of RIF resistance among the reported cases.

Study design and settings
This prospective study was carried out at Fayoum University Hospital, in collaboration with Fayoum Chest Hospital, Fayoum, Egypt. Patients enrolled in the current study were from both sexes with suspected PTB or EPTB, during the period from December, 2016 to December, 2019. In this study, we compared the PTB and EPTB detection capabilities of the GeneXpert assay, to bacterial culture on Lowenstein-Jensen (LJ) medium, and to composite reference standards (CRS). About 778 patients with highly suspected TB based on the clinical data; the relative laboratory tests results and the radiological findings; however, they did not start anti-tuberculosis treatment (ATT) yet at the time of registration, were included in this study. Excluded from this study were patients who were reported to have tuberculosis and started ATT; those who were unable to get proper samples for examination, patients refused or were unable to give written valid consent, in addition to patients with an underlying clinical diagnosis other than TB. Patients whose cultures grew as non-tuberculosis mycobacteria (NTM), those who were lost and/or died during follow-up, were banned from the current study.

Clinical specimen's collection and processing
Novel Research in Microbiology Journal, 2021 A total of 571(73.4 %) sputum specimens and 207(26.6%) extra-pulmonary specimens were included in this study. Sputum volume of at least 2-3 ml was considered as optimum and was processed for analysis. The smallest volumes of extra-pulmonary specimens required were as follows: 3 ml for any kind of body fluid including pus, 2.5 ml for cerebrospinal fluid (CSF); and 1 cm by 1 cm for biopsy specimens. After centrifugation of the sterile body fluids, the pellets were used. Non-sterile clinical specimens were processed by the conventional N-acetyl L-cysteine-NaOH (NALC-NaOH) method, for making smears, cultures and GeneXpert tests, according to Kawai et al., (2006);Zeka et al., (2011). The invasively collected specimens were processed directly.

Acid-fast bacilli smears and culture on Lowenstein-Jensen (LJ) medium
The processed specimens were used for microbiological examination. Sputum specimens were obtained from each patient, and were subjected to smear microscopy by ZN staining and culture on LJ media, following the protocol of Singh et al., (2016). Cultures were dealt with as a reference standard for measuring the accuracy of the GeneXpert assay as a diagnostic test.

GeneXpert MTB/RIF assay
The GeneXpert assay (Cepheid Inc, USA) was performed according to the manufacturer's instructions. In brief, expectorated sputum specimen (about 0.5 ml) and GeneXpert reagent were added in 1:2 ratio; vortexed twice and then incubated for 15 min. at room temperature until completely homogenized. About 2 ml of this mixture was pipetted into GeneXpert test cartridge, and then the cartridge was loaded into the GeneXpert machine. After 90 min., the GeneXpert system interpreted the results according to the measured fluorescent signals. According to Khadka et al., (2019), results were reported as negative if MTB was not detected, and considered as positive if MTB was detected; with or without rifampicin resistance. All samples that were culture positive (CP) and GeneXpert assay negative, and samples that were CN (culture-negative) and GeneXpert assay positive were retested twice, and the last result was considered for the study. Bias in reading GeneXpert results were minimized as it were interpreted by an independent observer who didn't know the results of CRS.

Composite reference standard for comparison
Culture on LJ medium is considered as an accepted reference standard, which is widely recognized as the best available and accurate method for isolation and detection of MTB. However, in paucibacillary diseases such as EPTB, growth of mycobacteria on this culture medium may be limited, as reported by Vadwai et al., (2011). The molecular tests including polymerase chain reaction (PCR) can detect DNA from dead bacterial cells with a limit of detection ranging from 5-100 bacilli/ ml; thus it may be used to identify CN samples. As a result of the seriously suboptimal reference standard of culture on LJ medium for EPTB, thus we also compared the GeneXpert results to a composite reference standard (CRS) assay; to test its true diagnostic potential for EPTB, in reference to Denkinger et al., (2014).The CRS may have poor specificity, hence, both CRS and culture on LJ medium were considered as reference standards to attain the best sensitivity and specificity. For PTB, diagnosis was used if any two of the following tests were positive: smear/ culture/ response to treatment/ radiological findings. Similarly, for EPTB, diagnosis was carried out if any two of the following assays were positive: smear/ culture/ histopathology/ cytology/ biochemical analysis/ response to treatment/ adenosine deaminase (ADA) levels; for sterile body fluids, pleural fluid, ascetic fluid, cerebrospinal fluid (CSF)/ and radiological findings (Vadwai et al., 2011). The patients were followed for 3 months after diagnosis and initiation of treatment. Improvement in signs and symptoms was considered as an adequate response to treatment. Meanwhile, if the case was improved on non-antituberculosis treatment (non-ATT), it was considered negative.
Novel Research in Microbiology Journal, 2021

Patients categories
According to the CRS and ATT follow-up, the patient's cases were categorized into four diagnostic groups by two experts, who were blinded to results of the GeneXpert test, in reference to Moussa et al., (2016). These four groups include: confirmed TB cases (CP, with SP or SN), probable TB cases (CN but has clinical symptoms, cytology/ histology and/or radiological findings, indicative of TB), possible TB cases (culture and other tests are negative with a clinical assumption of TB, and response to empirical ATT after 3 months follow-up), and no TB (all tests were negative for TB, and the case has improved without having ATT).

Statistical analysis
Statistical analysis was carried out with the SPSS for Windows (version 16.0). Numerical variables were summarized with mean ± SD (standard deviation). The significant differences among groups were assessed by the Student t-test, whereas analysis of categorical variables was examined by the chi-square test. Fisher Exact test was used for two by two tables if expected values were less than 5. A value of p≤ 0.05 was considered significant. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the GeneXpert assay for TB diagnosis were considered, and 95 % CI (confidence interval) was also calculated. The Kappa co-efficient was calculated to indicate level of agreement between readings of any two tests. It was described as follows: 0.2= none, 0.21-0.39= minimal, 0.40-0.59=weak, 0.60-0.79= moderate, 0.80-0.89=strong, 0.90-11=almost perfect (Ou et al., 2015).

Results
After exclusion of contaminated cultures (n= 12), insufficient specimens (n= 9), NTM positive specimens (n= 10), patients who were missing to follow-up (n= 22), and those who died (n= 6); about 778 samples (571 pulmonary specimens and 207 extrapulmonary specimens) from 778 patients were analyzed for the study. The socio-demographic data of the patients are described in Table (1). Patients aged between 4-88 years old, mean± SD is 44.3± 14.9 years. According to their residence, 447 (57.4 %) were living in rural areas, 321 (41.3 %) in urban areas, and 10 (1.3 %) were prisoners. Patients from rural areas significantly predominate over those from urban areas (p= 0.004). The extra-pulmonary specimens from 207 patients were variably distributed, as demonstrated in Table (2).

Ziehl-Neelsen smears microscopy
In the current study, ZN smear has detected MTB in pulmonary samples more often than in extrapulmonary specimens and the difference is statistically significant (p< 0.001) (

GeneXpert MTB/RIF assay
Out of the 778 specimens included in this study, the GeneXpert assay detected MTB in 152 (19.5%), compared to 144 (18.5%) specimens by culture, while 10 specimens gave errors (1.3 %) (    the GeneXpert assay is outlined in Table (2). No statistical difference is observed for detection of MTB in PTB and EPTB specimens by the GeneXpert assay (p=0.887), as clear in Table (1 Table (3).

Relative diagnostic efficiencies (overall sensitivity, specificity, PPV and NPV) of the GeneXpert assay, with reference to the LJ culture medium
With reference to the culture results, the sensitivity of the GeneXpert assay for the combined PTB and EPTB is 100 % (95 % CI. 72 %-100 %), the specificity is 98.7 (95 % CI. 97.8 %-99.6 %), PPV is 94.7 % (95 % CI. 90 %-97.3 %), and NPV is 100 % (95 % CI. 99 %-100 %). For pulmonary and extra-pulmonary samples the sensitivity, specificity, PPV, and NPV of the GeneXpert assay, using a culture as reference, is demonstrated in Table (4). The sensitivity of the GeneXpert assay is 100 % for CP-SP and CP-SN, in pulmonary and extrapulmonary specimens (Table 4). Statistically, on using the culture results as a reference standard, no significant difference is observed in the sensitivity of the GeneXpert assay between PTB and EPTB cases (p =1).

The inter-rater agreement of the GeneXpert assay and other tests for PTB and EPTB
The GeneXpert assay showed almost perfect agreement with culture on LJ medium for both pulmonary (kappa= 0.977) and extra-pulmonary specimens (kappa= 0.936), and for the combined samples (kappa= 0.967), as demonstrated in Table  ( 4). The agreement between the smear and culture is strong (kappa= 0.857) for pulmonary specimens, minimum (kappa= 0.389) for extra-pulmonary specimens, and moderate (kappa= 0.763) for the combined samples. According to the results presented in Table (5), the agreement between the GeneXpert assay and the culture on LJ medium, with the CRS for pulmonary specimens is strong (kappa= 0.826 and 0.804); respectively, although it is moderate for ZN staining (kappa= 0.733). This is remarkably better than those for extra-pulmonary specimens; where the GeneXpert assay and culture on LJ medium showed minimal agreement with the CRS (kappa= 0.358 and 0.340); respectively, while ZN staining showed no agreement with the CRS (kappa= 0.0).

Discussion
This study aimed to assess and compare the diagnostic precision of the GeneXpert assay for the PTB and EPTB. Overall, in the current study, the GeneXpert assay diagnosed more EPTB cases than the bacterial culture did; however, this difference was not statistically significant. In contrast, Bankar et al., (2018) recently reported a marked difference Novel Research in Microbiology Journal, 2021 in MTB detection between the bacterial cultures vs. the GeneXpert assay. A previous study conducted by Vadwai et al., (2011) attributed this lower positivity to the paucibacillary character of EPTB samples, and the tendency of MTB to form clumps with uneven distribution of bacilli cells. Meanwhile, Denkinger et al., (2014) reported that in diagnostic accuracy studies, an imperfect reference standard may lead to misclassification of samples. This agrees with the current study, where an assessment of the diagnostic accuracy according to two reference standards namely; the culture and a CRS, has provided a reasonable range for the sensitivity and specificity of the GeneXpert assay.
On using culture as the reference standard, results of the present study showed that the overall sensitivity, specificity, PPV, and NPV of the GeneXpert assay for the pulmonary samples were higher than other previous studies from China (Ou et al., 2015) and Ethiopia (Geleta et al., 2015). Moreover, the sensitivity for CP-SN was also higher than that recently reported by the study of Rasheed et al., (2019). This detected variation in sensitivities between the different studies could be attributed to the population studied, genetic differences, and the relative different prevalence of TB among the various populations.
Results of this study revealed that the GeneXpert assay is highly specific for the PTB diagnosis, in accordance with other earlier studies that have also reported this high specificity (Scott et al., 2011;Geleta et al., 2015). On the other hand, we currently recorded high specificity of the GeneXpert with regard to CN-SN pulmonary samples, in contrast to several recent studies which reported lower specificity among smear-negative PTB cases (Kawkitinarong et al., 2017;Ullah et al., 2017;Rasheed et al., 2019). This high specificity of GeneXpert assay suggests that it acts as a rapid test for the diagnosis of PTB in SN specimens, compared to the traditional methods in resource-limited settings.
In this study, with reference to the bacterial culture, the sensitivity of the GeneXpert assay was 100 % for SP-CP and SN-CP pulmonary samples. However, previous studies conducted by Boehme et al., (2010);Helb et al., (2010);Zeka et al., (2011);Armand et al., (2011) have reported similar sensitivity for SP-CP pulmonary specimens and lower sensitivity with SN-CP pulmonary ones. This difference between studies may be due to the quality of samples, and differences in the diagnostic gold standard.
According to Kohli et al., (2018), there is no specific recommendation for the use of the GeneXpert assay in specimens other than sputum. The sensitivity of the GeneXpert assay in EPTB observed in this study is similar to other previously published studies of Tortoli et al., (2012);Zmak et al., (2013); Sharma et al., (2014); Singh et al., (2016), which measured its sensitivity with reference to a bacterial culture. Among the CP specimens detected in this study, using the bacterial culture as the reference standard, the GeneXpert assay exhibited excellent sensitivity (100 %) for SP-CP and SN-CP extra-pulmonary specimens. These results correlate well with the previously published studies conducted by Zmak et al., (2013);Bankar et al., (2018).
In the current study, with reference to the CRS, the GeneXpert assay had high specificity, but limited sensitivity for MTB detection in extra-pulmonary specimens. Although the positive results can be of use in rapid identification of the disease; however, negative results offer less confidence for excluding EPTB. The GeneXpert assay had different sensitivities (33 % to 100 %) for MTB detection in EPTB in different types of specimens, similar to the recent study of Allahyartorkaman et al., (2019). Furthermore, with the CRS as the reference standard, the GeneXpert assay had high sensitivity for SP and low sensitivity for SN extra-pulmonary specimens, similar to the previous report of Armand With reference to the CRS, the sensitivity of the GeneXpert assay varied considerably between the extra-pulmonary specimen types. Its sensitivity for pleural effusion specimens was markedly lower than that for other extra-pulmonary specimens. Suboptimal performance of the assay was observed for body fluids; pleural effusion, CSF and urine, while its sensitivity was better with bone and LN biopsy, which is consistent with several previous studies of Tortoli et al., (2012); Patel et al., (2013); Sharma et al., (2014). As the available GeneXpert assay buffer has been assigned for sputum, the sensitivity for specimens other than sputum could give rise to many false-negative results, as reported by Theron et al., (2014).
The limited diagnostic utility of the GeneXpert assay in pleural fluid is attributed to the poor sensitivity of the assay in pleural fluid (Porcel, 2009;Ahmed et al., 2020). However, in the investigations of pleural TB, with available resources, the GeneXpert assay should be considered, since it has better sensitivity than staining and is more rapid than histology and bacterial culture, as highlighted by Denkinger et al., (2014). A recent work conducted by Tadesse et al., (2019) reported that we can't rely on a negative result of GeneXpert assay for the exclusion of the diagnosis of EPTB in fluid specimens, and thus ATT should be started in patients with a high clinical possibility of EPTB.
According to a meta-analysis study conducted by Denkinger et al., (2014) that assessed the diagnostic precision of the assay versus a CRS, the pooled sensitivity of GeneXpert with pleural fluid samples was 21.4 % versus 33.7 % in the present study. However, the GeneXpert sensitivity for CSF was similar to ours. Furthermore, and parallel to the current findings, Vadwai et al., (2011) reported low sensitivity of the GeneXpert assay for detecting the MTB in CSF. According to Denkinger et al., (2014), the WHO recommends the GeneXpert assay as the preferred initial test for the diagnosis of tuberculous meningitis.
Limited numbers of Egyptian studies have evaluated the diagnostic ability of the GeneXpert assay for PTB diagnosis, with reference to the bacterial culture (Moussa et al., 2016;Omar et al., 2019a;Tag Eldin et al., 2019). However, the tuberculosis pleural effusion was the only EPTB specimen tested by the GeneXpert assay in two recent Egyptian studies conducted by Omar et al., (2019b);Ahmed et al., (2020). Both studies have reported poor sensitivity but good specificity of the assay. The GeneXpert assay offers rapid detection of rifampicin resistance with reasonable precision (Singh et al., 2016). Rifampicin resistance was detected in only 13/152 (8.6 %) samples in this study, which was lower than that detected by Tag Eldin et al., (2019), but similar to that recorded by Omar et al., (2019a). In this study, the GeneXpert assay had a strong inter-rater agreement with the bacterial culture for both PTB and EPTB; respectively, which differs from the findings of Allahyartorkaman et al. (2019).
Finally, lack of studying the diagnostic precision of the GeneXpert assay on samples other than those assessed in this study (e.g., blood), in addition to shortage of studying the impact of the assay on patient's outcomes; are the most recognized limitations of this study.

Conclusion
The GeneXpert assay showed better sensitivity for the diagnosis of the PTB than the EPTB. Diagnosis of EPTB should be based on combining many tests such as bacterial culture and the GeneXpert assay. The diagnostic accuracy of this assay was high in ruling-in, but not in ruling-out of EPTB. The manufacturer or the WHO should provide standard recommendations for the nonrespiratory sample preparation.