Performance of Xpert® MTB/RIF among tuberculosis outpatients in Lilongwe, Malawi

Background Xpert® MTB/RIF is a molecular test for the detection of Mycobacterium tuberculosis and rifampicin resistance. It is considered to be a great advance over smear microscopy and culture. However, there is very little information regarding the performance characteristics of Xpert MTB/RIF in Malawi. Objective We aimed to evaluate the performance of Xpert MTB/RIF in a Malawian setting. Methods Stored sputum pellets were processed on Xpert MTB/RIF between June 2012 and May 2014. Results were compared to mycobacteria growth indicator tube and Löwenstein-Jensen cultures, LED fluorescent microscopy and GenoType® MTBDRplus assay. Rifampicin resistance was confirmed by DNA sequencing. Results Of the 348 specimens with valid Xpert MTB/RIF results, 129/348 (37%) were smear-positive and 198/348 (57%) were culture-positive. Xpert MTB/RIF demonstrated a sensitivity of 93.8% (95% CI 89.4% – 96.8%) and specificity of 97.4% (95% CI 93.5% – 99.3%), with a positive predictive value of 97.8% (95% CI 94.6% – 99.4%) and a negative predictive value of 92.6% (95% CI 87.4% – 96.1%). Xpert MTB/RIF correctly identified 185/186 (99.5%) rifampicin-sensitive and 2/2 (100%) rifampicin-resistant M. tuberculosis strains. Mutations were not detected by sequencing in one isolate which was rifampicin resistant on Xpert MTB/RIF but sensitive on MTBDRplus. Four non-tuberculous mycobacteria grew from four smear-negative specimens, namely, M. avium (n = 1) and M. intracellulare (n = 3). No cross-reactivity was observed with any of the non-tuberculous mycobacteria when using Xpert MTB/RIF. Conclusion When fully implemented, Xpert MTB/RIF may have an impact on patient care in Malawi. The increased diagnostic yield of Xpert MTB/RIF over smear microscopy can increase laboratory-confirmed tuberculosis detection and ensure that treatment is given to appropriate individuals or groups.


Introduction
the Malawi National Tuberculosis Programme (NTP) is using Xpert MTB/RIF to increase tuberculosis case detection and to detect rifampicin resistance as a proxy for MDR tuberculosis followed by culture (liquid and solid) and drug-susceptibility testing (DST) as confirmatory tests. 2 DST is not routinely performed in Malawi, with the exception of retreatment cases and individuals at risk for MDR tuberculosis. 3 This can lead to the development of underreporting of drug resistance, unnecessary suffering, additional costs for patients, and treatment with suboptimal regimens. 5 Consequently, this might increase the spread of drug-resistant tuberculosis in the population. 1 The Xpert MTB/RIF can both detect tuberculosis and identify rifampicin resistance in a single, rapid assay, 6,7 and is reported as having excellent sensitivity and specificity in detecting Mycobacterium tuberculosis and rifampicin resistance. 6,8 Some recent evaluations have demonstrated that Xpert MTB/RIF accurately detects 72.5% of smear-negative and 98.2% of smear-positive cases, and rifampicin resistance is detected with a specificity of 100% and sensitivity of 99.1%. 6,9 The test has an overall pooled sensitivity of 90.4% (95% CI 89.2% -91.4%), which is much higher than smear microscopy but is similar to that of solid culture in detecting tuberculosis. 1,7 Among smear-positive tuberculosis samples, Xpert MTB/RIF has a pooled sensitivity of 98.7% (95% CI 98.0% -99.2%), but has a substantially lower sensitivity of 75% (95% CI 72.0% -77.8%) among smear-negative tuberculosis samples. The overall pooled sensitivity in diagnosing for rifampicinresistant tuberculosis is 94.1% (95% CI 91.6% -96.0%) and 97.0% (95% CI 96% -97.7) pooled specificity. 7 Furthermore, use of Xpert MTB/RIF has been linked to improved diagnosis, resulting in early appropriate treatment. 10 The Malawi NTP recommended rolling out Xpert MTB/ RIF in 2012 to increase detection of tuberculosis, especially in sputum smear-negative and HIV-positive individuals. While Xpert MTB/RIF has been approved for clinical use, its optimal use in the clinical algorithm for Malawi has not been established. There are currently over 50 GeneXpert ® instruments distributed across 40 of the 315 public and private diagnostic centres (laboratories) with acid-fast bacilli testing capacity in the country. 2 Xpert MTB/RIF's performance is not known in the adult outpatient population in Malawi.
In this study, we evaluated the performance of Xpert MTB/ RIF in detecting M. tuberculosis complex and determining resistance to rifampicin among both HIV-positive and HIVnegative outpatients at Bwaila Hospital in Lilongwe, Malawi.

Laboratory methodologies
We used frozen stored sputum pellets (n = 351) obtained from an observational cohort study which collected 702 samples between April 2011 and July 2012 at Bwaila in Lilongwe, Malawi. The study was looking at the prevalence of drugresistant tuberculosis among adult outpatients (≥ 18 years) with laboratory-confirmed or clinically-diagnosed tuberculosis registering for tuberculosis treatment at this HIV/tuberculosis clinic. 11 Pellets were selected at random for use in this study without any special criteria to eliminate bias. Testing of these pellets using Xpert MTB/RIF started in June 2012 at the University of North Carolina (UNC) Project laboratory, Lilongwe, Malawi. Pellets were re-suspended in 1.5 mL phosphate buffer and an aliquot of 0.5 mL was processed on the Xpert MTB/RIF assay. In the primary study, all sample processes were performed at the UNC Project laboratory. In brief, sputum smears were prepared and stained with auramine-O stain. Smear microscopy was performed using LED fluorescent microscopy. Mycobacterial cultures (reference standard) were performed using both BACTEC mycobacteria growth indicator tube media (liquid) and Löwenstein-Jensen slants (solid). The reference standard was considered positive if there was growth of M. tuberculosis on either of the media and negative if both media were negative. Drug susceptibility was investigated using the GenoType ® MTBDRplus assay (version 2) (Hain Lifesciences GmbH, Nehren, Germany). Data collected from the above processes were included with data in this evaluation.

Xpert MTB/RIF
Samples for Xpert MTB/RIF testing were prepared as per manufacturer's instructions. Sample reagent was added to 500 µL of the re-suspended pellet in the ratio of 3:1 (sample reagent:specimen) as described previously. 12 Results were available within two hours of sample loading.

DNA extraction
Genomic bacterial DNA was extracted using a GenoLyse ® kit (Hain Lifescience GmbH, Nehren, Germany) from mycobacteria growth indicator tube liquid media by transferring 1.0 mL to a Sarstedt micro-centrifuge tube. The tube was centrifuged for 15 minutes at 14 000 rpm. The supernatant was carefully removed and the pellet was resuspended in 100 µL lysis buffer, vortexed thoroughly and incubated for five minutes at 95°C in a heating block. At the end of the five-minute incubation, tubes were briefly centrifuged to remove condensation. Neutralization buffer (100 µL) was added to each tube, vortexed for five seconds and then centrifuged for five minutes at 14 000 rpm. Supernatant was transferred to a new tube and the pellet discarded. Samples were stored at -80°C for use as DNA template. http://www.ajlmonline.org Open Access

Line probe assays
The GenoType ® MTBDRplus (version 2) (Hain Lifesciences GmbH, Nehren, Germany) was performed on positive culture isolates according to manufacturer's instructions to identify M. tuberculosis complex and to determine drug susceptibility to rifampicin and isoniazid as described previously. 13 Genotype ® CM (Hain Lifesciences GmbH, Nehren, Germany) was used to identify other common Mycobacterium species in samples which tested negative for M. tuberculosis complex on GenoType MTBDRplus.

PCR and DNA sequencing for the rifampicin resistance determining region
Determination of eligibility for DNA sequencing was based on detection of resistance to rifampicin by Xpert MTB/RIF. All eligible M. tuberculosis strains were shipped to the HIV Genotyping Laboratory (University of the Witwatersrand) and were processed as described below.

DNA amplification
Amplification of the 450 bp rpoB gene fragment which included the rifampicin resistance determining region was carried out by using forward primer rpoBF2 (5'-GAG GGT CAG ACC ACG ATG AC-3') and reverse primer rpoBR2 (5'-GAG CCG ATC AGA CCG ATG T-3') in a GeneAmp PCR system 9700 thermocycler (Applied Biosystems, Foster City, California, United States). The total volume of the reaction mix was 50 µL, which contained 5 µL of 10X high fidelity buffer, 1 µL of 10 mM dNTP mix, 2 µL of 50 mM MgSO 4 , 1 µL of each primer, 0.2 µL of Taq HiFi (Invitrogen, Carlsbad, California, United States), 36.8 µL of molecular grade water and 3 µL of genomic DNA. Amplification conditions were set as follows: 94°C for 2 minutes (initial denaturation); followed by 35 cycles of 94°C for 30 seconds, 55°C for 30 seconds, 68°C for 40 seconds; followed by a 10-minute final elongation at 68°C. Amplified products were visualised on a 1% agarose gel after staining with ethidium bromide.

Statistical analysis
All data were double entered into a Microsoft Excel spreadsheet (Microsoft, Redmond, Washington, United States). Discrepancies were checked against source records for completeness and consistency. Data analyses were done in Stata Version 12 (StataCorp, College Station, Texas, United States). The following statistics were calculated: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and 95% confidence intervals (CI). Mycobacterium culture was considered the reference standard.   D516V). It was observed that for both strains, the corresponding Xpert MTB/RIF probes gave cycle threshold (Ct) value = 0 and were also detected as resistant to rifampicin by GenoType MTBDRplus line probe assay. One (0.5%) smearnegative, culture-positive tuberculosis strain was resistant on Xpert MTB/RIF but sensitive on GenoType MTBDRplus (Table 3). A delayed amplification on probe B was observed in this strain, with a Ct value of 24.4. Further testing on the isolate using DNA sequencing revealed no rpoB gene mutation.

Discussion
In the present study, the Xpert MTB/RIF assay detected 58 more tuberculosis cases than smear microscopy. The paucibacillary nature of samples cultured from tuberculosis/HIV co-infected individuals revealed a more serious inability to detect tuberculosis using smear microscopy than was the case for Xpert MTB/RIF. Among HIV-positive individuals registering for tuberculosis treatment, a relatively larger percentage of those who were smear-negative, were also Xpert MTB/RIF and culture negative, and thus had no laboratory confirmation of disease. Overall analysis revealed expected results using the Xpert MTB/RIF assay, proving more sensitive and specific than fluorescent microscopy in detecting M. tuberculosis when using culture as the reference standard and there was no overlap of the confidence intervals. The sensitivity of Xpert MTB/RIF was lower in HIVpositive than in HIV-negative individuals. The Xpert MTB/RIF correctly distinguished M. tuberculosis from nontuberculous mycobacteria. Among the non-tuberculous mycobacteria, three were isolated from HIV-positive individuals, and one was from an HIV-negative individual.
Findings from our study are in agreement with studies by Rahman et al. 14 19,20 The possible explanation for this discrepancy could be differences in study design, considering that the current study used processed pellets while the other studies performed their evaluations using unprocessed sputum.  Prior studies have shown greater accuracy of Xpert MTB/RIF as compared to smear microscopy 6 and sensitivities of Xpert MTB/RIF ranging from 78% to 100% for smear-positive/ culture-positive samples. 21,22 Since Xpert MTB/RIF identified 58/219 (26.5%) of smear-negative samples, it can therefore assist in rapid detection of tuberculosis cases among smearnegatives and has the potential to impact patient care, although a small percentage (17%, 12/70) of the smearnegative/culture-positive cases was missed by Xpert MTB/ RIF. In Malawi, smear-negative adult tuberculosis patients have poor treatment outcomes with high death rates, largely due to concurrent HIV infection. 23 It is of particular interest to note that 64% of smear-negative samples which did not grow on culture and where M. tuberculosis was not detected by Xpert MTB/RIF, were from HIV-positive individuals. This may be interpreted as significant incorrect diagnosis of tuberculosis with a likelihood of treating people for tuberculosis who do not have the disease. It is therefore of paramount importance that steps should be taken to improve tuberculosis diagnosis to avoid exposing non-tuberculosis patients to the adverse effects of the tuberculosis medication. Misdiagnosis of tuberculosis can be a contributing factor to the high death rate observed in HIV-positive individuals globally. 1 Due to the increased diagnostic yield and short turn-around time of Xpert MTB/RIF as compared to culture, Xpert MTB/RIF offers an opportunity to improve the diagnosis of tuberculosis, both in HIV-positive and HIVnegative individuals, if used as a first diagnostic for all presumptive tuberculosis cases in Malawi, considering the high proportion of misdiagnoses observed in this study.
Our results are comparable to other studies which have shown that Xpert MTB/RIF increased tuberculosis case detection by almost 31%, despite its low sensitivity with smear-negative samples. 20 Another study demonstrated that 89% of rifampicinresistant tuberculosis patients started second-line treatment when Xpert MTB/RIF was introduced and that the average time to start second-line treatment was reduced from 1 -1.5 months by culture to one week with Xpert MTB/RIF. 23 The current study does not highlight this for Malawi since Xpert MTB/RIF testing was performed retrospectively and the initial study did not have a long-term follow-up on patients, as it was only concentrated on diagnostics. False resistance to rifampicin on Xpert MTB/RIF has been reported previously, although in some of the studies heteroresistance could not be excluded, because neither cartridge amplicons nor unprocessed sputum were sequenced, 17,25 which was also the case in our study. We performed DNA sequencing on isolates from culture, whereas Xpert MTB/ RIF was performed on sputum pellets. Theron et al. 22 7 In the present study, the sensitivity, specificity, PPV and NPV were not calculated due to the small number of rifampicin-resistant isolates identified.
The low rifampicin resistance detected in this study is similar to results obtained by Glynn et al., which identified 3/373 samples as resistant to rifampicin in Karonga, in the northern region of Malawi. 27 Low rifampicin resistance detection among incident cases observed in this study reflects well on the prevalence of rifampicin-resistant tuberculosis strains in communities surrounding Bwaila. However, it is debatable whether Xpert MTB/RIF results alone can be used for making a decision to start on second-line treatment in outpatients detected with resistance to rifampicin in Malawi, considering the number of isoniazid mono-resistances detected in previous studies. 11 It is therefore crucial to test tuberculosis strains for resistance to isoniazid as well, and to confirm rifampicin resistance detected by Xpert MTB/RIF before switching to second-line treatment. The success of the tuberculosis-control programme could be measured by the level of anti-tuberculosis drug resistance in a community. Consequently, the level of drug resistance gives future indications of suitable drug regimens. 27,28 The Malawi NTP has phased in plans within the context of national strategic plan for tuberculosis to expand and maintain Xpert MTB/RIF as a primary test for active case finding at high-volume antiretroviral therapy services and among high risk and vulnerable populations. This will consequently lead to improved tuberculosis diagnosis and initiation of appropriate treatment to patients, thereby reducing the spread of tuberculosis and/or transmission of resistant strains in the country. Additionally, Xpert MTB/RIF offers to help in reducing the time lost in the diagnostic pathway and potential loss of patients at each step of the pathway. 2 As part of the scale-up plan for the GeneXpert technology, the Malawi NTP plans to assess the concordance between phenotypic and genotypic technologies in order to inform interpretation of single rifampicin-resistance results. Systematic confirmation of MDR tuberculosis status of rifampicin-resistant patients diagnosed with Xpert MTB/RIF is lacking in the country. 2 Findings from the current study form a bank of data which can be beneficial to the NTP in its assessment of the Xpert MTB/RIF assay.

Limitations
With our limited sample size from a single outpatient location, the observed results may not apply across the country due to differences in the prevalence of HIV in rural districts. The relatively small number of patients with drug-resistant tuberculosis limits the ability to evaluate comparative performance of Xpert MTB/RIF. We recommend that future studies extend to other sites with an overall sample size and/or focus on evaluating Xpert MTB/RIF on raw sputum, since is it known that testing frozen stored specimens may influence results. 29 In the current algorithm for Malawi, a residual specimen from the second sample of presumptive tuberculosis patients is tested on Xpert MTB/RIF if both samples are smearnegative. Xpert MTB/RIF is also performed on samples from all patients suspected of having MDR tuberculosis and retreatment cases, but not on new smear-positive cases. 3

Conclusion
Our results demonstrate that Xpert MTB/RIF has the potential to increase the number of patients initiated early on appropriate treatment, given that additional tuberculosis cases were identified among smear-negative/cultureconfirmed cases, one of which was resistant to rifampicin. It is expected that the use of Xpert MTB/RIF will offer a greater opportunity in increasing tuberculosis case detection among high-risk and vulnerable groups in Malawi where sputum smear microscopy is still widely used. We believe that results obtained by Xpert MTB/RIF may facilitate a proactive approach to tuberculosis by enhancing decisions towards treatment and ensuring that the tuberculosis programme is not treating people who do not have tuberculosis but are obviously sick. When this is achieved, it is anticipated that there will be a reduction in the burden of tuberculosis among people living with HIV.