ABSTRACT
Early detection of resistance to second-line antituberculosis drugs is important for the management of multidrug-resistant tuberculosis (MDR-TB). The GenoType MTBDRsl version 2.0 (VER 2.0) line probe assay has been redesigned for molecular detection of resistance-conferring mutations of fluoroquinolones (FLQ) (gyrA and gyrB genes) and second-line injectable drugs (SLID) (rrs and eis genes). The study evaluated the diagnostic performance of the GenoType MTBDRsl VER 2.0 assay for the detection of second-line drug resistance compared with phenotypic drug susceptibility testing (DST), using the Bactec MGIT 960 system on Mycobacterium tuberculosis complex isolates from South Africa. A total of 268 repository isolates collected between 2012 and 2014, which were rifampin monoresistant or MDR based on DST, were selected. MTBDRsl VER 2.0 testing was performed on these isolates and the results analyzed. The MTBDRsl VER 2.0 sensitivity and specificity indices for culture isolates were the following: FLQ, 100% (95% confidence interval [CI] 95.8 to 100%) and 98.9% (95% CI, 96.1 to 99.9%); SLID, 89.2% (95% CI, 79.1 to 95.6%) and 98.5% (95% CI, 95.7 to 99.7%). The sensitivity and specificity observed for individual SLID were the following: amikacin, 93.8% (95% CI, 79.2 to 99.2%) and 98.5% (95% CI, 95.5 to 99.7%); kanamycin, 89.2% (95% CI, 79.1 to 95.6%) and 98.5% (95% CI, 95.5 to 99.7%); and capreomycin, 86.2% (95% CI, 68.3 to 96.1%) and 95.9% (95% CI, 92.2 to 98.2%). An interoperator reproducibility of 100% and an overall interlaboratory performance of 93% to 96% were found. The overall improvement in sensitivity and specificity with excellent reproducibility makes the GenoType MTBDRsl VER 2.0 a highly suitable tool for rapid screening of clinical isolates for second-line drug resistance for use in high-burden TB/HIV settings.
INTRODUCTION
The 2015 World Health Organization (WHO) Global Tuberculosis Report notes that approximately 3.3% of new cases of tuberculosis (TB) and 20% of previously treated TB cases in 2014 were multidrug-resistant TB (MDR-TB), defined as resistance to isoniazid (INH) and rifampin (RIF). South Africa is listed among the 14 countries that appear in all three WHO 2016 to 2020 revised high-burden country (HBC) lists (1). South Africa, with a high burden of MDR-TB, also experienced an increase in extensively drug-resistant TB (XDR-TB), accounting for 59% of XDR-TB patients reported globally in 2011 (1, 2). XDR-TB is defined as MDR-TB with additional resistance to any fluoroquinolone (FLQ) and at least 1 second-line injectable drug (SLID) among the aminoglycoside (AG) drugs amikacin (AMK) and kanamycin (KAN) and the cyclic peptide capreomycin (CAP) (2–5). Drug resistance in Mycobacterium tuberculosis isolates occurs at a low frequency due to spontaneous chromosomal mutations (6), which are selected through the improper use of anti-TB agents and low patient compliance with treatment, exerting selective pressure for the emergence of drug-resistant mutants (acquired resistance) (7).
In South Africa, despite the emphasis on TB therapy, the HIV epidemic has seriously hampered TB management and severely affected treatment outcomes (8), while the TB-HIV coepidemic has fueled the escalation of both MDR-TB and XDR-TB (9). The introduction of new rapid molecular diagnostic tests in South Africa, notably the Xpert MTB/RIF assay (Cepheid, USA) and line probe assays (LPAs), for the detection of drug-resistant TB has markedly improved patient management with decreased result turnaround times (TATs) for testing. Culture-based phenotypic drug susceptibility testing (DST), considered to be the gold standard for drug resistance determination, is important for MDR-TB confirmation and the assessment of drug resistance to second-line and new drugs in the management of MDR-TB and XDR-TB (10). However, conventional DST is labor intensive, time-consuming, and generally takes between 2 and 3 weeks of incubation to provide meaningful treatment directing results. In addition, second-line DST is fraught with challenges due to variability of methodology, reproducibility in performance, and reliability of results. Currently, there are no rapid genotypic diagnostic tests in the South African TB diagnostic algorithm for the determination of resistance to second-line anti-TB drugs (10). Molecular-based assays designed to detect specific drug resistance-encoding mutations in M. tuberculosis have the advantage of achieving faster TATs (within 48 h) for resistance reporting compared to conventional DST, and in the process it alerts clinicians to the emergence of drug resistance in M. tuberculosis strains from individual patients. Early detection of drug resistance is crucial to preventing the transmission of drug-resistant TB and averting mortality, as previously described (9, 11).
A current limitation of molecular assays is that they do not accommodate all mutations conferring resistance to anti-TB agents. A WHO Expert Group determined in 2013 that the GenoType MTBDRsl version 1.0 (VER 1.0) (Hain LifeScience, Germany) cannot replace phenotypic DST, but it may be used as a rule-in test for XDR-TB (12). In a meta-analysis published in 2014, Theron and colleagues (5) reported respective pooled sensitivity and specificity indices of 83.1% (95% confidence interval [CI], 78.7 to 86.7%) and 97.7% (95% CI, 94.3 to 99.1%) for FLQs and 76.9% (95% CI, 61.1 to 87.6%) and 99.5% (95% CI, 97.1 to 99.9%) for SLID for culture isolates using the GenoType MTBDRsl VER 1.0 assay. They concluded that since the assay only targets selected mutations involving gyrA (FLQ) and rrs (SLID) gene loci, mutations encoding resistance to FLQ and SLID that occur outside these regions would be missed by the assay (5).
GenoType MTBDRsl VER 2.0 is redesigned based on GenoType MTBDRsl VER 1.0 and accommodates additional mutations for the molecular detection of resistance to FLQ involving gyrA and gyrB and SLID resistance covering both rrs and eis genes (13). The probes target commonly occurring mutations that encode resistance to these agents. The gyrA probes target codons 85 to 97 of the gene, and rrs probes target nucleic acid positions 1401 to 1484. The inclusion of additional targets for selected mutations in the gyrB region (codons 536 to 541) and eis promoter region (−10 to −14) for low-level KAN resistance are reported to improve the performance of the assay for the detection of FLQ and SLID resistance (13).
In order to prioritize and facilitate the identification of pre-XDR-TB (resistance to INH, RIF, and to any FLQ or SLID but not both) (8) and XDR-TB in South Africa's high-burden setting, the implementation of GenoType MTBDRsl VER 2.0 would be a fundamental improvement to case detection and management. An evaluation of this assay was undertaken with the objective of assessing diagnostic performance as well as to determine interoperator and interlaboratory performance of GenoType MTBDRsl VER 2.0 in the detection of second-line drug resistance mutations in M. tuberculosis complex culture isolates compared to phenotypic DST.
RESULTS
A total of 268 M. tuberculosis complex isolates were included in the study, and the frequency of resistance to OFX, KAN, AMK, and CAP among these isolates was detected by the two methods (Table 1). Of these, 42 isolates from the NHLS Braamfontein Laboratory had phenotypic DST performed for OFX and KAN. As the DNA for the 42 isolates was not available, whole-genome sequencing (WGS) could not be performed. Resistance to one or more of the drugs under evaluation was observed in 120/268 (44.8%) of the M. tuberculosis isolates included in the study. All 27 probe bands of GenoType MTBDRsl VER 2.0 strips were interpretable in all samples tested with successful positive and negative quality controls (14). Phenotypically susceptible isolates were correctly classified as susceptible by the assay.
Summary of GenoType MTBDRsl VER 2.0 LPA and DST results
Detection of FLQ resistance.Sensitivity, specificity, and accuracy indices for FLQ (OFX) resistance using the GenoType MTBDRsl VER 2.0 assay were determined as 100%, 98.9%, and 99.3%, respectively, compared to the gold standard (Table 2). There were 85/120 (70.8%) isolates resistant to FLQ as tested by DST. GenoType MTBDRsl VER 2.0 detected 92 mutations in the gyrA and gyrB genes among the 85 FLQ (OFX)-resistant isolates. The distribution of mutations is summarized in Table 3. The majority of mutations, 52/92 (56.5%), were observed at codon 94. Other gyrA mutations detected by the assay were at codon 90 (29/92; 31.5%) and at codon 91 (9/92; 9.8%).
Performance of GenoType MTBDRsl VER 2.0 assay
Hybridization patterns observed in strains classified as resistant by LPA using GenoType MTBDRsl VER 2.0 assaya
The diversity of single defined mutations at gyrA codon 94 included the following (Table 3): gyrA MUT3C (D94G), 23/52 (44.2%); gyrA MUT3A (D94A), 7/52 (13.5%); gyrA MUT3D (D94H), 6/52 (11.5%); and gyrA MUT3B (D94N/D94Y), 5/52 (9.6%). In 3/52 (5.8%) isolates, both gyrA mutations MUT3B and MUT3D (D94N/D94Y and D94H) were detected. Other gyrA defined mutations detected at codons 90 and 91 were the following: gyrA MUT1 (A90V), 25/29 (86.2%); and gyrA MUT2 (S91P), 7/9 (77.8%) (Table 3).
Heteroresistance in gyrA was observed in 14/85 (16.5%) OFX-resistant isolates. Eight isolates among these exhibited two mutation bands involving codons 94, 91, and 90. Results for these isolates were the following: 4/85 (4.7%) wild-type (WT)/gyrA MUT1 and MUT3B (A90V and D94N/D94Y), 2/85 (2.4%) WT/gyrA MUT2 and MUT3B (S91P and D94N/D94Y), and 2/85 (2.4%) WT/gyrA MUT3B and MUT3C, respectively (Table 3).
There were 8/85 (9.4%) isolates with undefined mutations that were interpreted as resistant. The following wild-type bands were not detected among these isolates: gyrA WT2, 4/85 (4.7%); gyrA WT3, 1/85 (1.1%); and gyrB WT1, 3/85 (3.3%). WGS performed on 6/8 available isolates confirmed defined mutations at gyrA (A90V, S91P, and D94Y) missed by the assay and gyrB mutations at codons 274 and 499 that are not covered by the assay (Table 4). All 8 of these isolates were phenotypically resistant to OFX. An isolate that harbored gyrA MUT3C (D94G) mutation confirmed by WGS with an A94G mutation was phenotypically susceptible to OFX. In the selection of isolates tested, no defined mutations were detected by the assay in the gyrB region (i.e., mutations N538D and E540V).
WGSa results for MTBDRsl VER 2.0 isolates with undefined mutations and discordant DST
Detection of SLID resistance.The sensitivity, specificity, and accuracy of GenoType MTBDRsl VER 2.0 for SLID resistance was 89.2%, 98.5%, and 96.3%, respectively, compared with phenotypic DST. The respective individual SLID sensitivity and specificity assessments were the following: 93.8% and 98.5% for AMK, 89.2% and 98.5% for KAN, and 86.2% and 95.9% for CAP (Table 2).
GenoType MTBDRsl VER 2.0 detected 56 defined mutations and 6 undefined mutations in either the rrs or eis genes among the 65/120 (54.2%) isolates phenotypically resistant to SLID (Table 3). The most frequently observed mutation (48/56; 85.7%) for SLID resistance was rrs MUT1 (A1401G), with four of these isolates displaying heteroresistance. All 48 isolates were confirmed as phenotypically resistant to a SLID (AMK, KAN, or CAP). Two isolates had rrs WT1 bands absent with no corresponding mutation band observed and 1/2 tested was wild type by WGS, with both being phenotypically susceptible to SLID. Similarly, there were 4/65 (6.2%) isolates that were interpreted as resistant (absence of eis WT2 and no mutation band observed), 3/4 were phenotypically resistant to KAN, and 1/4 was phenotypically susceptible to KAN. WGS tested 2/4 isolates as wild type, with a C-10T/G-10A mutation observed in 1/4 isolates. The eis MUT1 (C-14T) mutation was observed in 8/65 (12.3%) of the SLID-resistant isolates and confirmed phenotypically as KAN resistant. In 6/65 (9.2%) isolates phenotypically resistant to KAN, no rrs or eis mutation was observed. These were interpreted as susceptible to SLID by the assay. DNA available for 3/6 of these isolates tested wild type by WGS (Table 4). Among five isolates (5/65; 7,7%) that were phenotypically susceptible to CAP, 2/5 had the rrs MUT1 (A1401G) mutation and 3/5 had eis missing WT2, and they were interpreted as resistant to SLID by the assay. WGS performed on these isolates revealed no resistance-conferring mutations for gidB and tlyA regions. The mutation rrs MUT2 (G1484T) was not observed among any of the isolates tested.
Detection of XDR-TB.Among the isolates tested, 31/120 (25.8%) were XDR by phenotypic DST (Table 5). Agreement between GenoType MTBDRsl VER 2.0 and phenotypic DST was calculated as 97.0% for the detection of XDR-TB. GenoType MTBDRsl VER 2.0 correctly identified 27/31 (87.1%) isolates, missing 4/31 (12.9%) isolates, all of which were phenotypically SLID (KAN) resistant but sensitive by the assay. WGS was performed for 2/4 isolates missed by the assay, and they were wild type for rrs and eis; however, both isolates were phenotypically resistant to KAN (SLID). All isolates with a defined mutation showed phenotypic resistance.
MGIT DST and GenoType MTBDRsl VER 2.0 analysis of XDR isolates identified (n = 31)
Reproducibility.Among the 5 laboratories that tested the panel of 30 isolates, the overall performance ranged between 93% and 96% (Table 6). The interoperator assessment showed 100% agreement.
Reproducibility assessment of participant laboratories for GenoType MTBDRsl VER 2.0
DISCUSSION
To our knowledge, the evaluation of GenoType MTBDRsl VER 2.0 in conjunction with a reproducibility assessment is the first such study conducted in a high TB-HIV setting in Africa. GenoType MTBDRsl VER 2.0 has shown an improvement in sensitivity and specificity for the determination of molecular resistance to both FLQ (100% [95% CI, 95.8 to 100%] and 98.9% [95% CI, 96.1 to 99.9%]) and SLID (89.2% [95% CI 79.1 to 95.6%] and 98.5% [95% CI 95.7–99.7%]) compared with the pooled sensitivity and specificity of GenoType MTBDRsl VER 1.0 reported by the WHO Expert Group (12) and a meta-analysis by Theron and colleagues (5). Agreement between the phenotypic gold standard and GenoType MTBDRsl VER 2.0 for FLQ (OFX) and SLID was 99.3% and 96.3%, respectively, across a wide distribution of mutations. The assay was found to be highly reproducible in terms of interlaboratory (93 to 96%) and interoperator (100%) assessment.
FLQ resistance in M. tuberculosis is ascribed mainly to gyrA mutations, with 57.5% of mutations detected at codon 94 and 31.5% at codon 90 (15). Consistent with published data, the highest frequency of mutations conferring FLQ resistance was observed in gyrA codon 94, followed by codon 90 (Table 3). Furthermore, gyrA mutations D94G and D94N were observed in the majority of the isolates tested. Previous studies have indicated that isolates harboring these mutations exhibit high levels of resistance to FLQ (15, 16). The gyrA MUT3D/D94H mutation, reported as a rare in silico mutant (13, 17), was observed in 9/85 (10.6%) of mutations identified from these clinical isolates. Heteroresistance involving either FLQ or SLID was detected in 18/85 (21.2%) of isolates, with multiple mutation bands observed in some isolates (Table 3), and were confirmed as phenotypically resistant. Cohen et al. reported on mixed-strain infections in a hospital setting in KwaZulu-Natal, South Africa (18). Further investigation is required to elucidate the frequency and clinical relevance of mixed infections across a more widespread South African setting. Based on our findings, interpretation of these strains as resistant was consistent with the phenotypic DST results.
Undefined mutations with the absence of gyrA wild-type bands were confirmed as phenotypically resistant. WGS results available for four isolates confirmed the presence of gyrA mutations that the assay missed. The GenoType MTBDRsl VER 2.0 guideline for interpretation of results states that “only bands with intensities as strong as or stronger than the amplification control zone (AC) are to be considered” (13). Three out of four of these discordant isolates had gyrA MUT2/S91P bands with intensities less than that of the AC zone and therefore were interpreted as negative and reported as resistant due to the absence of gyrA wild-type bands. No weak-intensity mutation bands were detected in the fourth isolate. An isolate phenotypically susceptible to OFX with gyrA MUT 3C/D94G bands and a confirmed Asp94Gly mutation on WGS had gyrA WT bands with less intensity than the AC control band. The discrepant OFX phenotypic DST is possibly due to a mixed population. The inclusion of selected gyrB probes offered limited enhancement to the identification of FLQ mutations, as only the absence of gyrB WT was observed in 3/268 (1.1%) isolates. GenoType MTBDRsl VER 2.0 is limited to detecting mutations in selected areas of the quinolone resistance-determining regions (QRDR) of gyrA and gyrB only; therefore, FLQ resistance mechanisms outside these regions are likely to be missed but do not appear to be a major concern in our setting presently.
Consistent with published data (17), rrs MUT1 A1401G (translating to high-level SLID resistance) was the most frequently observed mutation (77.5%) among tested isolates. The overall SLID sensitivity has improved in the new assay to 89.2% compared with previously published data (5, 12). The inclusion of eis promoter region probes improved the detection of SLID resistance, as observed in 8/62 (12.9%) low-level KAN-resistant isolates, all phenotypically KAN resistant and not associated with rrs mutations. Two isolates, with rrs WT1 bands absent, no corresponding mutation band observed, wild type by WGS, and interpreted as resistant to SLID by the assay, were phenotypically susceptible to SLID. The omission of WT bands in the line probe assay is not a reliable indication of phenotypic resistance and would require confirmatory phenotypic DST (19). In 6 phenotypically KAN-resistant isolates that were AMK susceptible, no mutations were detected by the assay or WGS. Although other factors related to phenotypic resistance cannot be excluded, the discrepant results may be due to the inherent challenges of in-house (nonstandardized) second-line drug preparations available for testing (19). Five isolates that were phenotypically susceptible to CAP, interpreted as SLID resistant by the assay and wild type by WGS, displayed resistance only to KAN and AMK phenotypically. As indicated by Georghiou et al. (20), rrs MUT1 A1401G mutation is a moderate predictor of CAP resistance. Therefore, CAP resistance should be confirmed phenotypically prior to exclusion of the drug from a treatment regimen (21).
There were 31 isolates identified as phenotypically XDR, and the GenoType MTBDRsl VER 2.0 assay achieved 97.0% agreement between the index and gold standard tests. Of the four XDR cases misclassified by the assay, all showed FLQ resistance but missed SLID resistance. All were phenotypically KAN resistant. One isolate was classified as XDR by LPA with gyrA MUT1 (A90V) and a missing eis WT2 only (C-12T/G-10A) but was phenotypically susceptible to KAN. We used strict criteria, and resistance to any one SLID was accepted as a criterion to classify a strain as XDR. Due to limited availability of standardized drug preparations for second-line DST, in routine practice, laboratories would normally only test one drug out of this class for resistance determination. This is a limitation of the data available for this sample set, as mutations in the assay are known to correlate well with OFX resistance and is evident in our observations. As reported, correlation of these mutations with higher FLQs is uncertain and may still show phenotypic susceptibility despite the presence of these mutations; thus, phenotypic testing remains important for these drugs (22, 23).
The reproducibility assessment of selected routine laboratories showed excellent performance consistent with other data on molecular diagnostics (Table 6) (24). GenoType MTBDRsl VER 2.0 LPA performed well with excellent reproducibility and high sensitivity and specificity for FLQ and SLID resistance determination. The average reporting turnaround times for GenoType MTBDRsl VER 2.0 LPA varied from 2 to 4 days subsequent to the identification of a positive TB culture, whereas second-line phenotypic DST results were only available in 14 to 21 days. The longer TAT for DST results is largely dependent on the viability/fitness of the isolate, especially for XDR isolates. It also highlights the importance of direct testing in providing clinically relevant and reliable results.
Furthermore, adoption of the assay in laboratories already performing the GenoType assays was relatively easy and quick to implement. GenoType MTBDRsl VER 2.0 LPA is therefore suitable for testing clinical isolates as a rapid screening tool for detection of second-line drug resistance in countries with a high burden of MDR-TB.
A limitation of the study was that only culture isolates were used and not clinical specimens, as this study was performed to compare directly the GenoType MTBDRsl VER 2.0 assay with the gold standard, phenotypic DST. However, there is a need for further studies evaluating the GenoType MTBDRsl VER 2.0 assay on clinical specimens from patients with MDR-TB, thereby enhancing the turnaround time for resistance reporting and pre-XDR/XDR case detection in high-risk settings. Another limitation was that of the FLQs, only OFX was tested and none of the other later-generation FLQs (isolates collected between 2012 and 2014). This probably explains the higher sensitivity found in our study compared with other studies (25). Phenotypic DST for later-generation FLQs (moxifloxacin) was introduced in 2015 per WHO recommendations (22, 23). A general limitation of second-line drug testing is the limited availability of standardized drug preparations (19). A further limitation was that samples from the Braamfontein TB Referral Laboratory could not be included for WGS resolution testing, as the DNA was not available.
In a press release by the WHO in May 2016, the use of the GenoType MTBDRsl VER 2.0 assay as “an initial test, instead of phenotypic culture-based DST” to detect FLQ and SLID resistance in confirmed RIF-R and MDR patients was recommended (26, 27). Appropriately trained laboratory staff, quality assurance, and availability of laboratory infrastructure are requisite recommendations of the WHO to implement use of this assay (26, 27). Our study provides support and evidence for these recommendations and the implementation of the assay in South Africa.
MATERIALS AND METHODS
Setting and study design.The evaluation of the GenoType MTBDRsl VER 2.0 assay was conducted at the Centre for Tuberculosis, National Institute for Communicable Diseases (CTB, NICD), Johannesburg, South Africa. The laboratory is a designated WHO Supranational Laboratory (SRL) accredited to ISO 15189:2012 (14). The Human Research Ethics Committee of the University of Witwatersrand, Johannesburg, South Africa, approved the study (M150752). The study was structured with two interrelated components involving the validation of characterized M. tuberculosis isolates and a reproducibility assessment using a subset of these isolates tested by selected National Health Laboratory Service (NHLS) laboratories in South Africa, performing diagnostic tests involving culture of M. tuberculosis.
Mycobacterial isolates.A total of 268 repository isolates collected between 2012 and 2014 were tested. These were comprised of 92 phenotypically well-characterized M. tuberculosis complex isolates using whole-genome sequencing (WGS) and 176 anonymized M. tuberculosis clinical isolates exhibiting rifampin monoresistance (RIF-R) or MDR based on phenotypic DST, tested at the CTB or NHLS TB Referral Laboratory, Braamfontein, Johannesburg, South Africa. Phenotypic testing for later generations of FLQ and SLID was introduced in 2015 (2). These isolates originated from Gauteng, Limpopo, Northern Cape, North West, KwaZulu-Natal, and Mpumalanga provinces, collected during surveillance activities for rifampin resistance in South Africa. The fully susceptible ATCC M. tuberculosis H37Rv 27294 reference strain (http://www.atcc.org/Products/All/27294.aspx ) was used as the DST and LPA positive quality control culture.
Phenotypic DST.Bactec MGIT 960 DST using EpiCenter software (Becton, Dickinson and Company Diagnostic Systems, Sparks, MD, USA) for interpretation of results was performed according to the manufacturer's recommendations (28) and considered the gold standard for resistance determination. The following critical concentrations of drugs recommended by WHO for testing of drug-resistant TB using Bactec MGIT 960 DST were used: ofloxacin (OFX), 2.0 μg/ml; AMK, 1.0 μg/ml; KAN, 2.5 μg/ml; and CAP, 2.5 μg/ml (22).
GenoType MTBDRsl VER 2.0.The GenoType MTBDRsl VER 2.0 assay was performed according to the manufacturer's instructions (13). Each strip contains 27 reaction zones with probes for all specific targeted regions. Seven probes for gyrA (A90V, S91P, D94A, D94N/Y, D94G, and D94H) and 2 probes for gyrB (N538D and E540V) were used to detect FLQ resistance. SLID resistance was detected by selected rrs (A1401G, C1402T, and G1484T) and eis (C-14T and C-12T) probes. The presence of all wild-type bands and absence of mutation bands indicated susceptibility. The development of specific mutation bands (defined mutation) or the absence of wild-type bands (undefined mutation) (29) related to a specific gene on the hybridization strip was interpreted as resistance to the respective drug (13). Heteroresistance was demonstrated when both wild-type and/or mutation band(s) were present and was recorded as resistant for interpretation purposes. SLID resistance referred to resistance to at least one of the 3 injectable drugs (AMK, KAN, and CAP) (5).
Discordance resolution by whole-genome sequencing.WGS was used to resolve discordance between phenotypic and genotypic methods. WGS was performed using the MiSeq platform (Illumina, San Diego, CA). Library preparation was performed using the Illumina Nextera XT library preparation kit and sequencing reaction using the MiSeq version 3 cartridge (2 × 300 bp). Variant detection was performed using the resequencing module with the reference strain M. tuberculosis H37Rv (NC_000962 ) on CLC Genomics Workbench (v7.5.1).
Reproducibility.In order to evaluate the interlaboratory performance of the assay, a panel of 10 M. tuberculosis complex external quality assurance (EQA) isolates, constituting a subset of the isolates evaluated in the present study comprised of 10 isolates present in triplicate, was distributed to 5 NHLS laboratories performing diagnostic culture of M. tuberculosis as well as LPA (30). An overall agreement of ≥90% was deemed an acceptable score. Interoperator reproducibility was performed by comparing findings on 45 EQA isolates tested by two operators at CTB, working independently.
Statistical analysis.Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), based on agreement between the DST gold standard and the index assay and 95% confidence intervals (95% CIs), were calculated for both the antibiotic class and individual drugs. Agreement between the gold standard and index assay was calculated using the McNemar test. Statistical analyses were performed using Stata release 13 (31).
ACKNOWLEDGMENTS
We thank Hain LifeScience, South Africa, for providing GenoType MTBDRsl VER 2.0 reagents and training at participating laboratories, the staff members of CTB NICD/NHLS, and the participant NHLS laboratories for their assistance during the study.
Funding from the CTB NICD/NHLS supported this study.
We have no conflicts of interest to declare.
FOOTNOTES
- Received 8 September 2016.
- Returned for modification 9 November 2016.
- Accepted 9 December 2016.
- Accepted manuscript posted online 14 December 2016.
- Copyright © 2017 American Society for Microbiology.