Performance of the Xpert MTB/RIF assay for the diagnosis of pulmonary tuberculosis and rifampin resistance in a low-incidence, high-resource setting

Performance of the Xpert MTB/RIF assay, designed to simultaneously detect Mycobacterium tuberculosis complex (MTBC) and rifampin (RIF) resistance, has been well documented in low-resource settings with high TB-incidence. However, few studies have assessed its accuracy in low TB incidence settings. We evaluated the performance of Xpert MTB/RIF using clinical sputum specimens routinely collected from suspect pulmonary TB patients over a 4-year time period in San Diego County, California. Xpert MTB/RIF results were compared to acid-fast bacilli (AFB) smear microscopy, mycobacterial culture, and phenotypic drug susceptibility testing (DST). Of 751 sputum specimens, 134 (17.8%) were MTBC culture-positive and 2 (1.5%) were multidrug-resistant (MDR). For the detection of MTBC, Xpert MTB/RIF sensitivity was 89.6% (97.7% and 74.5% in smear-positive and -negative sputa, respectively) and specificity was 97.2%; while AFB smear sensitivity and specificity were 64.9% and 77.8%, respectively. Xpert MTB/RIF detected 35 of 47 smear-negative culture-positive specimens, and excluded 124 of 137 smear-positive culture-negative specimens. Xpert MTB/RIF also correctly excluded 99.2% (121/122) of nontuberculous mycobacteria (NTM) specimens, including all 33 NTM false-positives by smear microscopy. For the detection of RIF resistance, Xpert MTB/RIF sensitivity and specificity were 100% and 98.3%, respectively. Our findings demonstrate that Xpert MTB/RIF is able to accurately detect MTBC and RIF resistance in routinely collected respiratory specimens in a low TB-incidence setting, with comparable performance to that achieved in high-incidence settings; and suggest that under these conditions the assay has particular utility in detecting smear-negative TB cases, excluding smear-positive patients without MTBC disease, and differentiating MTBC from NTM.

Our aim was to evaluate the accuracy of the Xpert MTB/RIF assay in the diagnosis of pulmonary TB and RIF resistance in a low TB-incidence setting, as applied in the course of routine clinical practice. The results of this study have the potential to clarify the role of this application in TB clinical management, and to inform future testing protocols in low-incidence, high-resource settings.

Materials and methods
This study was conducted at the San Diego County Public Health Department, using data and specimens routinely collected by the microbiology section of the San Diego County Public Health Laboratory (SDCPHL). SDCPHL receives and processes clinical specimens from the county TB Clinic as well as other local medical treatment facilities, and routinely performs AFB smear microscopy, mycobacterial culture, phenotypic DST, and NAA testing. In cases of suspected DR TB, SDCPHL also routinely sends specimens to the California Department of Public Health (CDPH) Microbial Diseases Laboratory (MDL) for pyrosequencing (PSQ).
San Diego County has an annual TB case rate of 7.3 per 100,000, with 234 TB cases reported in 2015 [42]. Annually, approximately 18% of reported TB cases are resistant to at least one first-line anti-TB drug, and 1.1% have MDR TB [43,44].

Study design and specimens
In this retrospective study, all sputum specimens collected from patients in San Diego County that were tested using the Xpert MTB/RIF assay at the SDCPHL from Jan 2012 to Jan 2016 were included in the study. Specimens were collected from patients at the San Diego County TB Clinic and other county medical treatment facilities, in the course of routine clinical evaluations for pulmonary TB. At the SDCPHL, in addition to Xpert MTB/RIF testing each specimen underwent routine AFB smear microscopy and mycobacterial culture; and isolates from each MTBC culture-positive specimen underwent phenotypic DST. All corresponding test results were available for each specimen, and no specimens were excluded from the study.
Other variables obtained for each specimen included: patient age, dates of collection and reporting, and method of specimen collection (induced vs expectorated sputa). In addition, TB Control surveillance data on treatment status were obtained for all Xpert-positive culture-negative samples, and results of PSQ performed by the CDPH MDL were obtained for all culturepositive samples with RIF resistance detected by phenotypic DST or Xpert MTB/RIF.

AFB smear microscopy
Specimens were processed using the N-acetyl-L-cysteine-NaOH (NALC-NaOH) method for digestion and decontamination. Specimens were concentrated by centrifugation at 3,200 Xg for 20 min, and sediments were reconstituted with approximately 2.0 ml of 0.067 M sterile phosphate buffer (pH 6.8). Smear microscopy was performed on processed sediments using auramine-O fluorochrome staining, and confirmed with Ziehl-Neelsen (ZN) staining. Smearpositive specimens were graded from 1+ to 4+ according to CDC guidelines [6].

Mycobacterial culture
Processed sputum sediments from each sample were inoculated into Middlebrook 7H10 solid slants, as well as liquid broth Mycobacterial Growth Indicator Tubes (MGIT; Becton Dickinson [BD], Sparks, MD) using the BD BACTEC MGIT 960 automated system. Liquid and solid cultures were incubated at 37˚C for up to 6 and 8 weeks, respectively. Positive mycobacterial isolates were confirmed with ZN staining, and species identification was performed with conventional biochemical tests, observation of colony morphology, and/or nucleic acid hybridization using molecular probes (Accuprobe; Gen-Probe, San Diego, CA). Niacin accumulation and nitrate reduction biochemical identification tests were performed on Lowenstein-Jensen (LJ) solid culture media, inoculated from 7H10 or MGIT isolates with positive mycobacterial growth.

Xpert MTB/RIF assay
The Xpert MTB/RIF assay was run on the GeneXpert Dx instrument system according to the manufacturer's recommendations [27]. Briefly, after digestion, decontamination and concentration, 0.5 ml of re-suspended sediment was transferred to a conical screw-capped tube, 1.5 ml of Xpert MTB/RIF sample reagent was added by sterile pipette, and the tube was recapped and shaken vigorously 10-20 times. The sample was incubated for a total of 15 minutes at 20-30˚C, with manual agitation 10-20 times at one point between 5 and 10 minutes into the incubation period. The reagent-treated sample was then transferred by sterile pipette into the sample chamber of the Xpert MTB/RIF cartridge and loaded into the GeneXpert Dx instrument system for sample processing. In the event of "no result", "invalid" or "error" results, the test was repeated according to the manufacturer's recommendations using a new Xpert MTB/RIF cartridge.
The SDCPHL and TB Control Program employed the Xpert MTB/RIF assay in accordance with CDC recommendations for NAA testing: to perform on at least one respiratory specimen, preferably the first collected, from each patient with signs and symptoms of pulmonary TB and for whom the test result would alter case management or TB control activities [16,49]. Some variation in protocol existed during the 4-year study time period due to resource availability, however in general the Xpert MTB/RIF assay was used to reflexively test the first AFB smear-positive specimen collected from each patient, as well as AFB smear-negative specimens per clinician request in patients with a high suspicion of pulmonary TB.

Statistical analysis
Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and positive and negative likelihood ratios (LRs) were calculated for the detection of MTBC and RIF resistance using culture and phenotypic DST as the reference standards, respectively, with exact Clopper-Pearson 95% confidence intervals. Final test results were available for all specimens, and all were included in the analyses. Standard descriptive statistics were used to characterize the study population. Statistical analyses were performed using IBM SPSS Statistics, version 23.0 (IBM Corp., Armonk, NY) and GraphPad Prism, version 6.0 (GraphPad Software, La Jolla, CA).
This study was reviewed and certified as exempt by the institutional review board of the University of California San Diego, and all samples were accessed anonymously. In addition the study was reviewed and received approval from the County of San Diego Health and Human Services Strategic Planning and Operational Support Division.

Detection of rifampin resistance
Of the 2 specimens with discrepant Xpert MTB/RIF and phenotypic DST results, DNA sequencing revealed 1 specimen to have an rpoB gene mutation (CTG!CCG Leu511Pro) reported by the CDPH MDL to be associated with low-level but clinically significant RIF resistance. In the other discrepant isolate sequencing detected a silent rpoB mutation (TTC!TTT 514Phe) that is not known to confer RIF resistance. For the 2 specimens with concordant Xpert MTB/RIF and phenotypic DST RIF r results, sequencing confirmed RIF resistance-conferring mutations (TCG!TTG Ser531Leu) in both isolates. Incorporating sequencing results for discrepancy resolution, Xpert MTB/RIF specificity and PPV increased to 99.2% (116/117) and 75% (3/4), respectively ( Table 4). The Xpert MTB/RIF assay also reported RIF resistance in 1 culture-negative specimen. This was one of the 15 Xpert-positive culture-negative specimens collected from patients on anti-TB treatment, and review of surveillance data revealed a silent rpoB mutation reported on a previously sequenced isolate from this patient.
Among all MTBC culture-positive specimens, phenotypic DST identified a total of 16 specimens with resistance to any first-line anti-TB agent, as detailed in Table 5

Discussion
Accuracy of the Xpert MTB/RIF assay for the diagnosis of pulmonary TB and RIF resistance assay has been previously described, primarily in studies conducted in resource-poor countries of the world with high TB incidence [28][29][30][31][32]. In respiratory specimens the assay has demonstrated a pooled sensitivity of 89% (98% and 67% among smear-positive and-negative specimens, respectively) and specificity of 99% for the detection of MTBC, and pooled sensitivity and specificity of 95% and 98% for the detection of RIF resistance, respectively [32]. In this study we evaluated the performance of Xpert MTB/RIF in routine clinical practice over a 4-year time period in San Diego County, a low TB-incidence setting with an annual case rate of 7.3 per 100,000 [42]. Our results compare favorably with pooled data from global studies, and demonstrate the assay's ability to perform equally well in a low TB-incidence setting.
Xpert MTB/RIF and other NAATs have consistently shown high sensitivity among smearand culture-positive specimens, however their variable sensitivity in smear-negative specimens has raised concern [16,19]. This test characteristic is of particular importance in low TB-incidence settings, where it is estimated that up to 60% of TB cases are smear-negative [33]. An Xpert MTB/RIF study performed in Canada (4.6 cases per 100,000) reported a sensitivity of 28% in smear-negative sputum specimens, and suggested the assay may have limited utility in low-incidence, high-resource countries [41]. In the present study, despite an MTBC prevalence of only 8.9% (47/527) among smear-negative specimens, Xpert MTB/RIF correctly detected 74.5% (35/47) of smear-negative culture-positive specimens, with a specificity of 99.2%. This performance was superior to the pooled meta-analyses, and suggests that for the detection of smear-negative MTBC, the assay has utility in both high-and low TB-incidence settings.
Results from our study illustrate the marked improvement in case detection that Xpert MTB/RIF offers over AFB smear microscopy (sensitivity and specificity of 64.9% and 77.8%, respectively). Of 134 culture-positive specimens in this study, 47 (35.1%) were smear-negative. Among those MTBC specimens missed by smear microscopy, 35 were correctly detected by Xpert MTB/RIF, increasing overall case detection by 26.1% (35/134). Perhaps more importantly for U.S. TB programs to consider, however, is that microscopy falsely identified 137 MTBC culture-negative specimens, of which 124 (90.5%) were correctly excluded by Xpert MTB/RIF. In this setting the assay appears to have a significant advantage over AFB smear microscopy, not only in detecting smear-negative TB cases, but also ruling out smear-positive individuals without MTBC disease.
NAATs, Xpert MTB/RIF included, have proven especially useful in settings with high NTM-burden, as they are able to rapidly differentiate MTBC from NTM and potentially avoid unnecessary treatment [19,50,51]. This is relevant in the U.S. where NTM-exceeds TB-burden, and evidence suggests that NTM infections are increasing by approximately 8% annually, with clinically significant pulmonary NTM infections increasing in particular [52,53]. The prevalence of NTM in our study was 16.3% (122/751), of which Xpert MTB/RIF correctly excluded 99.2% (121/122). In contrast, AFB smear microscopy falsely identified 33 NTM specimens as MTBC. These false-positives represented 24.1% (33/137) of smear-positive culture-negative specimens, and Xpert MTB/RIF correctly excluded them all. With these performance characteristics, the assay appears to have particular utility in this low TB-incidence setting with high NTM-burden.
The Xpert MTB/RIF assay reported 14 false-negative and 17 false-positive results in this study, referent to culture. Of the 14 Xpert-negative culture-positive specimens, 12 were smearnegative and 2 were 1+ smear-positive. As has been described previously, this can likely be attributed to low bacillary load, given that the Xpert MTB/RIF reported limit of detection (LOD) in sputum (131 cfu/mL) is higher than culture (10-100 cfu/mL) [6,24]. Of the 17 Xpert-positive culture-negative specimens, 15 were collected from patients on anti-TB drugs at the time of testing. The Xpert MTB/RIF assay is not intended for use in monitoring response to treatment, as it has the potential to detect genetic material from remnant, non-viable organisms [26,27]. This was likely the cause of the false-positives observed in our study. In clinical practice, situations may prompt clinicians to use the assay in this manner (eg., upon collection of a smear-positive specimen while monitoring treatment response in an established, previously culture-converted TB case); and given the intent of this study was to evaluate the assay's performance under routine practice conditions, these specimens were included. However, our findings demonstrate its limited utility in these circumstances, and reinforce that Xpert MTB/ RIF results must be questioned when testing patients on anti-TB drugs. Additionally, secondary analysis with exclusion of these specimens revealed substantially improved specificity (99.7%) and PPV (98.4%), in particular among smear-positive sputa (specificity and PPV of 100%), illustrating the assay's superior performance when used as recommended [15,27,49].
Consistent with the incidence of MDR TB in the U.S. and San Diego County, in our study phenotypic DST identified MDR TB in 1.5% (2/134) of MTBC isolates, which included all RIF r strains identified [43,44]. Xpert MTB/RIF detected both RIF r (MDR) specimens, and correctly identified 116 RIF s specimens. Referent to phenotypic DST, for the detection of RIF resistance (and consequently MDR TB), Xpert MTB/RIF sensitivity and specificity were 100% and 98.3%, respectively. In this study, in a setting with a low incidence of RIF resistance and MDR TB, these results compare favorably with global pooled meta-analysis findings.
Prior research has raised concern over the specificity of Xpert MTB/RIF for the detection of RIF r strains relative to phenotypic DST. However, given accumulating evidence that the reference standard may fail to detect some low-level but potentially clinically relevant RIF resistance, it appears that Xpert MTB/RIF specificity may be underestimated [11,15]. This evidence comes from a number of studies comparing solid and liquid culture-based phenotypic DST with rpoB gene sequencing, demonstrating that MTBC strains with specific rpoB mutations (e.g., CTG!CCG Leu511Pro, GAC!TAC Asp516Tyr, CAC!CTC His526Leu, CAC!AAC His526Asn, CTG!CCG Leu533Pro) almost always appear susceptible on MGIT DST. Of note, while the majority of these "disputed" mutations are located within the RRDR of the rpoB gene (codons 507 to 533), some of the most common (Leu511Pro and Leu533Pro) are located at the extreme ends of the RRDR [11][12][13][14]. Silent RRDR mutations that do not result in an amino acid change and do not confer resistance have also been documented (eg., TTC!TTT 514Phe), however these appear to be rare [11,14,54,55]. Reviewing an original study on Xpert MTB/RIF performance [28], Van Deun et al. note that sequencing revealed RRDR mutations in all 9 discordant phenotypically susceptible strains, of which only 1 silent mutation was truly a false-positive, since all others were low MIC mutations not detectable on MGIT DST [11].
In the present study, of the 2 specimens with discrepant results (RIF r by Xpert MTB/RIF and RIF s by phenotypic DST), sequencing revealed 1 isolate with a commonly missed and potentially clinically relevant mutation (CTG!CCG Leu511Pro), and the other with a silent mutation (TTC!TTT 514Phe). Using sequencing for discrepancy resolution, Xpert MTB/RIF specificity for the detection of RIF resistance (and consequently MDR TB) increased from 98.3% to 99.2%, and PPV increased from 50% to 75%. Evidence suggests that while the assay has high NPV in all settings (>99%), PPV is dependent on the prevalence of RIF resistance, ranging from <70% to >90% in settings where the underlying prevalence is <5% and >15%, respectively [15]. Our results are compatible with these estimates, given the low prevalence of RIF resistance in this study, and historically in San Diego County (<2%) [56]. Considering the potential for false-positives and low PPV with Xpert MTB/RIF, our findings reinforce CDC recommendations that all RIF r results prompt rapid DNA sequencing for first-and secondline anti-TB drugs, in order to tailor MDR treatment to the patient's specific resistance profile [49].
In the U.S., working under general CDC guidance for NAA testing, TB programs have been challenged to determine the most effective and efficient use of Xpert MTB/RIF in their practice settings. The assay has been incorporated into a variety of testing algorithms (e.g., testing smear-positive specimens only vs smear-positive and high-suspicion smear-negative specimens vs testing specimens from all suspected cases), yet no one approach has gained consensus. In a study of Xpert MTB/RIF cost-effectiveness in the U.S., Choi et al. proposed that the addition of the assay would result in lower total health care costs compared to algorithms using microscopy and culture alone, and that testing at least one sputum specimen from all patients with signs/symptoms of TB would be more cost-effective than testing only smear-positives [57]. In the present study Xpert MTB/RIF was used to reflexively test AFB smear-positive specimens, as well as AFB smear-negative specimens in patients with a high suspicion of pulmonary TB. While this study did not evaluate costs associated with its use, our results suggest that in a low incidence TB-setting the assay can be highly effective when applied under this protocol.
Limitations of the present study include its retrospective design, lack of clinical data, and relatively low numbers of RIF r specimens. Information on treatment status was obtained for all patients with Xpert-positive culture-negative specimens, however it would be beneficial to incorporate further clinical data on all patients, allowing direct assessment of the assay's impact on diagnostic and therapeutic decision-making. It would also be useful to evaluate the effect of Xpert MTB/RIF testing on other TB control efforts, such as contact tracing and respiratory isolation precautions, as well as the associated financial costs and benefits. Research in these areas will help to inform future clinical and programmatic use of the assay, and have the potential to impact broad public health strategies for TB control.
In summary, our results demonstrate that in a low TB-incidence setting and under routine practice conditions the Xpert MTB/RIF assay is able to accurately detect MTBC and RIF resistance in clinical sputum specimens, with performance equal to that seen in high-incidence settings. It provides the ability to rapidly diagnose or exclude pulmonary TB disease, differentiate MTBC from NTM, and detect suspected MDR TB; with markedly improved case detection over smear microscopy, particularly in low TB-incidence settings with high NTM-burden. Additionally, this study suggests that in a low TB-incidence setting, Xpert MTB/RIF may be effectively employed within a protocol for testing smear-positive and high-suspicion smearnegative sputum specimens.