new england

journal of medicine
established in 1812 september 9, 2010 vol. 363  no. 11

Rapid Molecular Detection of Tuberculosis
and Rifampin Resistance
Catharina C. Boehme, M.D., Pamela Nabeta, M.D., Doris Hillemann, Ph.D., Mark P. Nicol, Ph.D.,
Shubhada Shenai, Ph.D., Fiorella Krapp, M.D., Jenny Allen, B.Tech., Rasim Tahirli, M.D., Robert Blakemore, B.S.,
Roxana Rustomjee, M.D., Ph.D., Ana Milovic, M.S., Martin Jones, Ph.D., Sean M. O’Brien, Ph.D.,
David H. Persing, M.D., Ph.D., Sabine Ruesch-Gerdes, M.D., Eduardo Gotuzzo, M.D., Camilla Rodrigues, M.D.,
David Alland, M.D., and Mark D. Perkins, M.D.

A bs t r ac t

Global control of tuberculosis is hampered by slow, insensitive diagnostic methods, From the Foundation for Innovative New
Diagnostics, Geneva (C.C.B., P.N., M.D.P.);
particularly for the detection of drug-resistant forms and in patients with human im- Forschungszentrum Borstel, Borstel, Ger-
munodeficiency virus infection. Early detection is essential to reduce the death rate many (D.H., S.R.-G.); the Department of
and interrupt transmission, but the complexity and infrastructure needs of sensitive Clinical Laboratory Sciences, University
of Cape Town, and National Health Labo-
methods limit their accessibility and effect. ratory Service, Cape Town (M.P.N., A.M.),
Methods and the Unit for Clinical and Biomedical
TB Research, South African Medical Re-
We assessed the performance of Xpert MTB/RIF, an automated molecular test for search Council, Durban (J.A., R.R.) — all
Mycobacterium tuberculosis (MTB) and resistance to rifampin (RIF), with fully integrated in South Africa; P.D. Hinduja National
Hospital and Medical Research Centre
sample processing in 1730 patients with suspected drug-sensitive or multidrug-resis- (Hinduja), Mumbai, India (S.S., C.R.); In-
tant pulmonary tuberculosis. Eligible patients in Peru, Azerbaijan, South Africa, and stituto de Medicina Tropical Alexander von
India provided three sputum specimens each. Two specimens were processed with Humboldt, Universidad Peruana Cayeta-
no Heredia, Lima, Peru (F.K., E.G.); Spe-
N-acetyl-l-cysteine and sodium hydroxide before microscopy, solid and liquid culture, cial Treatment Institution, Baku, Azer­
and the MTB/RIF test, and one specimen was used for direct testing with micros- baijan (R.T.); the Division of Infectious
copy and the MTB/RIF test. Diseases, New Jersey Medical School,
University of Medicine and Dentistry of
Results New Jersey, Newark (R.B., D.A.); Cepheid,
Among culture-positive patients, a single, direct MTB/RIF test identified 551 of 561 Sunnyvale, CA (M.J., D.H.P.); and the De-
partment of Biostatistics and Bioinfor-
patients with smear-positive tuberculosis (98.2%) and 124 of 171 with smear-nega- matics, Duke University Medical Center,
tive tuberculosis (72.5%). The test was specific in 604 of 609 patients without tu- Durham, NC (S.M.O.). Address reprint
berculosis (99.2%). Among patients with smear-negative, culture-positive tubercu- requests to Dr. Boehme at the Founda-
tion for Innovative New Diagnostics, Ave.
losis, the addition of a second MTB/RIF test increased sensitivity by 12.6 percentage de Budé 16, 1202 Geneva, Switzerland, or
points and a third by 5.1 percentage points, to a total of 90.2%. As compared with at
phenotypic drug-susceptibility testing, MTB/RIF testing correctly identified 200 of
This article (10.1056/NEJMoa0907847) was
205 patients (97.6%) with rifampin-resistant bacteria and 504 of 514 (98.1%) with published on September 1, 2010, at NEJM
rifampin-sensitive bacteria. Sequencing resolved all but two cases in favor of the .org.
MTB/RIF assay.
N Engl J Med 2010;363:1005-15.
Conclusions Copyright © 2010 Massachusetts Medical Society.

The MTB/RIF test provided sensitive detection of tuberculosis and rifampin resis-
tance directly from untreated sputum in less than 2 hours with minimal hands-on
time. (Funded by the Foundation for Innovative New Diagnostics.)
n engl j med 363;11  september 9, 2010 1005
The New England Journal of Medicine
Downloaded from on March 25, 2013. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.

which provides results analyses were performed by a statistician who within 2 hours. . text of this article at NEJM. HIV coinfection. uses heminested real-time polymerase. Patients in the group at risk for pul- tients leads to increased mortality. with symptoms suggestive of pulmonary tuber- portionate frequency of smear-negative disease in culosis or multidrug-resistant tuberculosis who HIV-associated tuberculosis. none of which were involved in the design before transferring a defined volume to the car. FIND authors In accordance with recommendations on was not involved in data collection. sitive case detection or drug-susceptibility test. and statistical into the GeneXpert device. and symptomatic contacts of structure restrict the use of such tests to reference patients with known multidrug-resistant disease. Azerbaijan. the Foundation for Innovative New Diag- tridge containing all reagents required for bacte. Me thods mated 500. All rights reserved. Additional development support rial lysis. or conduct of the study. those with tection of tuberculosis and multidrug-resistant tu.5 ml. whereas the group at risk for complexity of mycobacterial culture and current multidrug-resistant disease included patients who nucleic acid–amplification technologies for the de. which integrates sample This study was designed and supervised by the processing and PCR in a disposable plastic car. South Africa. receiving therapy. 2013. The study was (RIF). Xpert MTB/ The study protocol was reviewed and approved RIF. and ongoing transmission. Copyright © 2010 Massachusetts Medical Society.10. by eight institutional review boards or technical rium tuberculosis (MTB) and resistance to rifampin committees at the ministerial level. India.11 Test- ing is carried out on the MTB/RIF test platform Study Design and Oversight (GeneXpert. Peru. able at NEJM. conducted in accordance with the protocol (avail- chain-reaction (PCR) assay to amplify an MTB.) ration in a public–private partnership. and Mumbai. Study participation did with molecular beacons for mutations within the not alter the standard of care. (For details regarding the sites. All au- sign and conduct of diagnostic accuracy assess­ thors vouch for the accuracy and completeness of ments. and amplicon detection.7 The in the past 60 days. Baku. Cape Town and Durban. an automated molecular test for Mycobacte. aggravated by the dispro. prospective the data on March 25. secondary re.6. We enrolled consecutive adults ing. was provided by the National Institutes of  september 9.13 we undertook a multicenter.12 The only manual step Cepheid. worldwide each year have access to sufficiently sen. nostics (FIND). evaluation of the MTB/RIF test to determine its sensitivity and specificity in the intended target Laboratory Methods population as compared with the best available Patients meeting the clinical eligibility criteria were reference standard. and multidrug re- high-burden countries. laboratories. least 1.11  nejm.1 Diagnostic delay.9 a fully automated molecu. wrote the first draft of the manuscript. nonconverting pulmonary tuberculosis who were berculosis8 and the need for the associated infra. For personal use only. asked to provide three sputum specimens over a 1006 n engl j med 363. had not received a tuberculosis medication with- berculosis). and the Bill and Melinda Gates Founda- is the addition of a bactericidal buffer to sputum tion. The n e w e ng l a n d j o u r na l of m e dic i n e O nly a small fraction of the esti. Written informed consent was specific sequence of the rpoB gene.000 patients who have multi- drug-resistant tuberculosis and 1. Study Population lion patients who have coinfection with tuberculosis From July 2008 through March 2009. sponsor. which is probed obtained from all patients. we conduct- and the human immunodeficiency virus (HIV) ed this study at five trial sites in Lima. sistance. amplification. Cepheid). 2010 The New England Journal of Medicine Downloaded from nejm. No other uses without monary tuberculosis were eligible only if they sistance (including extensively drug-resistant tu. The MTB/RIF cartridge is then inserted investigators at each study site. available with the full resistance testing was developed through collabo. rifampin-resistance determining region. Supplementary Appendix. see the lar test for tuberculosis case detection and drug.2-5 The were able to provide three sputum samples of at failure to quickly recognize and treat affected pa.37 mil. is common. had undergone previous treatment. All patients were enrolled from populations that To respond to the urgent need for simple and were selected for diversity in the prevalence of rapid diagnostic tools at the point of treatment in tuberculosis. nucleic acid extraction. Data were collected by tridge.

org on March 25. indetermi- shipped to the German National Reference Labo.5 ml of phosphate buffer and subjected ratory results and clinical information by clinical to microscopy with Ziehl–Neelsen staining. Bias was minimized MGIT [mycobacteria growth indicator tube] 960 through blinding. and then were  september 9. Durban. n engl j med 363. For personal use only. smears of scanty grade (1 to 10 acid-fast bacilli per type MTBDRplus assay (Hain Lifescience) per. For three sites. a nega- ratory in Borstel for testing 1 to 5 days after col. culture. radiography.14 diagnosis for patients undergoing repeat analyses followed by centrifugation. A three sites. those with smear-negative. At and radiologic findings (clinical tuberculosis). agnosis was required.16 with the latter medium review of clinical records and were available for used only in Durban) and liquid medium (BACTEC only a subgroup of patients. 100 fields) or one or more smears of 1+ or more formed from culture isolates (in Baku) or from the (10 to 99 bacilli per 100 fields). culosis clinicians. and the MTB/ molecular and reference tests were not aware of RIF test. nate phenotypic rifampin susceptibility. review committees did not have access to nucleic The first positive culture from each specimen acid–amplification test results. treated for tuberculosis on the basis of clinical according to the manufacturer’s instructions. Because a clear final di- manufacturer’s instructions. since technicians performing culture. nary tuberculosis. The third sputum sample was tested the results of other tests. only Repeat tuberculosis analyses (smear. ysis if there was a negative culture result while the assured reference laboratories. In a random fashion. BD Microbiology Systems). case was defined as positive results on at least ple (in Cape Town and Durban). nate diagnosis were excluded from the main anal- All participating laboratories were quality. and review committees composed of three local tuber- cultivation on solid medium (egg-based Löwen. patients with suspected multidrug resistance). growth after collection. .e.and culture-negative Cape Town and Mumbai) or ProbeTec ET MTB for pulmonary tuberculosis who nonetheless were Complex Direct Detection Assay (BD) (in Baku). tuberculosis species nators received lists of patients for follow-up but by MPT64 antigen detection (Capilia TB. those with DNA that was extracted from the NALC–NaOH no bacteriologic evidence of tuberculosis who had centrifugation pellet of the first sputum sample improvement without treatment (no tuberculosis). except that smear. pected cross-contamination of cultures (i. A culture-positive NALC–NaOH pellet of the second sputum sam.and culture-positive pulmo- and Mumbai).11  nejm. The final l-cysteine and sodium hydroxide (NALC–NaOH). Detection of Tuberculosis and Rifampin Resistance 2-day period (two spot samples and one obtained other nucleic acid–amplification test was positive in the morning) (Fig. and clinical workup) days to growth in MGIT or <20 colonies in Löwen- were performed in patients who had smear. according to the one of four culture vials. Categories for Analysis stein–Jensen medium (for sites in Lima. All rights reserved. was established on the basis of conventional labo- pended in 1. 2010 1007 The New England Journal of Medicine Downloaded from nejm. or sus- lection. Tauns not the reasons for follow-up. two or if the patient was selected by the central data- of the three samples were processed with N-acetyl.. drug-resistant genotyping was carried smear-positive case was defined as at least two out by line-probe assay with the use of the Geno. HIV results were obtained by stein–Jensen15 or 7H11. base as a random control for follow-up. Copyright © 2010 Massachusetts Medical Society. patients with an indetermi- negative specimens were also tested.and stein–Jensen medium) or if the patient died or was culture-negative samples if the MTB/RIF test or lost to follow-up. The interpretation of directly by Ziehl–Neelsen microscopy and the data from MTB/RIF tests was software-based MTB/RIF test without NALC–NaOH decontami. with the use of Cobas Amplicor MTB (Roche) (in and those who were smear. and independent of the user. 1). one of four cultures had positive results after >28 MTB/RIF test. All study coordi- underwent confirmation of M. No other uses without permission. 2013. Sputum samples from Baku were of nontuberculous mycobacteria only. Laboratories)17 and indirect drug-susceptibility testing with the proportion method on Löwen. tive culture with a positive sputum smear. conventional nucle. Patients were divided into four categories for analy­ and Baku) or MGIT SIRE18 (for sites in Cape Town sis: those with smear. Clinical teams and nation. cul- ic acid–amplification testing was carried out on ture-positive pulmonary tuberculosis. con- rollment clinic and tested samples within 2 days tamination of at least three of four cultures. Study laboratories patient was receiving tuberculosis treatment (for for four sites were located within 5 km of the en.

and the second result was ment at enrollment because of suspected multidrug used for analysis. 105 (7. patients with complicated tuberculosis inactivated material was transferred to the test and a rate of multidrug resistance as high as 50% were cartridge (equivalent to 0.20 (Fig.2%) had clini- Statistical Analysis cally defined tuberculosis.  september 9. LJ denotes Löwenstein–Jensen.and culture-pos- ABI sequence-analysis software.8%) had smear. which reports a high rate of multidrug resistance (25%) among patients was located in the microscopy room or another with tuberculosis and a rate of HIV coinfection of ap- general-purpose laboratory space. The MTB/RIF test was performed as described Patients were enrolled at centers that have diverse previously19. test and the combination of three tests. Sequencing Bidirectional sequencing was performed on the ing equations were used for calculating confi- 81-bp rpoB core region of culture isolates in all ri. and NALC–NaOH system to the central database. Of pa- were estimated for a single direct test. version 5. groups and testing methods.5 ml of decontaminated pellet). Enrollment and Outcomes. The additional sample reagent in pellets Cape Town and Durban. tum or 0. Azerbaijan. No other uses without permission. Copyright © 2010 Massachusetts Medical Society. which screening and treatment facility. including 115 who had culture-negative indeterminate and sufficient sample remained. 207 of 741 on a pelleted sample. In pellets. the combination of two tests (27. Peru. a single test tients with culture-positive samples. clustering. A to- of three tests (one direct and two pelleted). Traces were analyzed with analysis. sputum samples (in 10). 1). In Baku. patients were was necessary to meet the volume requirements enrolled at primary care tuberculosis clinics located for the assay sample.22 ward (CGTGGAGGCGATCACACCGCAGAC) and reverse (AGCTCCAGCCCGGCACGCTCACGT) pri­ R e sult s mers with the use of the BigDye Terminator Cycle Sequencing kit. Sensitivity and Specificity fidence intervals. Sample re. a rate of coinfection with the human immunodeficiency virus (HIV) of less tum and in a 3:1 ratio to decontaminated sputum than 3%. dence intervals to account for within-patient fampin-resistant and discordant strains with for. enrolled at 30 primary care clinics with a high rate of agent was added in a 2:1 ratio to untreated spu.9%) had smear-negative. tuberculosis case notification.” or “no result” divided by were excluded from the analysis for a variety of the total number performed. The closed sputum container within informal settlements with a high incidence of was manually agitated twice during a 15-minute tuberculosis and an estimated rate of HIV coinfection of 70% and a rate of multidrug resistance of 4%. For personal use only.7 ml of untreated spu.11  nejm. 567 (38. A total of 268 patients sified as “invalid. The ples (in 103 patients) or an inadequate volume of indeterminate rate was the number of tests clas. The electronic proximately 6%. and a low rate of multidrug resistance. according to the manufacturer’s Patients recommendations. itive tuberculosis. tal of 113 patients were not eligible for testing binations were classified as positive if at least one because of an inadequate number of sputum sam- of the component test results was positive. generalized estimat. enrolled at a tertiary care center. Two laboratory technicians populations with a high prevalence of tuberculosis.” “error. South Africa. For analyzing the single direct resistance (Fig. the overall sensitivity of the MTB/RIF test was 1008 n engl j med 363. India. The n e w e ng l a n d j o u r na l of m e dic i n e MTB/RIF Test Figure 1 (facing page). and for comparisons across sub.0. culture-positive tuberculosis.2.1%) had no Sensitivity and specificity for the MTB/RIF test clinical evidence of tuberculosis (Table 1).21 For all other intrapatient Case Detection MTB/RIF results. In were trained as operators and passed proficiency Lima. Wilson’s binomial method was used to calculate 95% con. and 616 (42. patients with suspected tuberculosis were testing after four runs per person. and the combination on conventional drug-susceptibility testing. in a 3130xl Genetic Analyzer Of the 1462 patients (4386 samples) included in the (Applied Biosystems). Among patients with culture-positive tuberculosis. 174 (11.9%) were found to have multidrug resistance (one direct and one pelleted). 2013. N-acetyl-l-cysteine and sodium hydroxide. Car. When results were reasons. 2010 The New England Journal of Medicine Downloaded from nejm. All rights reserved. 2).org on March 25. prisoners were enrolled on arrival at a tuberculosis tridges were inserted into the test platform. the samples but were receiving tuberculosis treat- assay was repeated once. . In period at room temperature before 2 ml of the Mumbai. MGIT results were sent directly from the MTB/RIF test mycobacteria growth indicator tube.

Detection of Tuberculosis and Rifampin Resistance 1843 Patients were screened Inclusion criteria: Suspicion of pulmonary tuberculosis (no previous therapy) or multidrug-resistant disease (therapy failure. No other uses without permission.5 ml 113 Were not eligible for testing 103 Did not have 3 sputum samples 10 Had samples with insufficient volume 1730 Patients were eligible Sputum 1 Sputum 2 Sputum 3 NALC–NaOH NALC–NaOH Smear Smear Smear MGIT and LJ MGIT and LJ MTB/RIF Second molec- MTB/RIF ular test MTB/RIF 268 Were excluded 115 Had culture-negative suspected multidrug resistance while receiving therapy 28 Had contamination of ≥3 of 4 cultures 23 Had growth of nontuberculous mycobacteria only 10 Had indeterminate phenotypic rifampin result 39 Had smear-positive sample with all cultures negative 7 Had suspected culture cross-contamination 46 Died or were lost to follow-up 1462 Were included in the main analysis 741 Were culture-positive 721 Were culture-negative 567 Were smear-positive 105 Had clinical tuberculosis 174 Were smear-negative 616 Did not have tuberculosis n engl j med 363. For personal use  september 9. 2013. . All rights on March 25. Copyright © 2010 Massachusetts Medical Society.11  nejm. retreat- ment. or contact with multidrug-resistant patient) Age ≥18 yr Provision of informed consent Collection of 3 sputum samples >1. 2010 1009 The New England Journal of Medicine Downloaded from nejm.

2% for smear-nega. RIF test was 93.02). culture-positive cases.and and 90. a sample of 2 to 3 ml is transferred to the test cartridge.6%. Assay Procedure for the MTB/RIF Test.11 nejm. which is then loaded into the instru- ment.2% and nificant difference in sensitivity between tests on rose to 96. 85.5% for one test. The n e w e ng l a n d j o u r na l of m e dic i n e 1 Sputum liquefaction and inactivation with 2:1 sample reagent 4 5 6 7 Sample Ultrasonic lysis DNA molecules Seminested automatically of filter-captured mixed with dry real-time filtered and organisms to PCR reagents amplification washed release DNA and detection in integrated reaction tube 2 8 Transfer of Printable 2 ml material test result into test cartridge 3 Cartridge inserted into MTB-RIF test platform (end of hands-on work) Time to result.8% for smear. The mixture is shaken.9%. 1 hour 45 minutes Figure on March 25.” PCR denotes polymerase chain reaction.0% with the additional testing of a pel.2% for three tests. 97.6% for two MTB/RIF tests. The sensitivity was 99. the sensitivity of the The estimated specificity was september 9.1% for two tests. untreated sputum and those on decontaminated leted sample. direct MTB/RIF test. 2010 The New England Journal of Medicine Downloaded from nejm. 98.16).2% for a single assay was 72. RIF resistance not detected. A single. pellet (P = 0. RIF test identified a greater proportion of culture- tive. culture-positive tuberculosis. 2013. Among HIV-positive patients with pul- imens per patient had a modest effect over the monary tuberculosis. For personal use only. such as “MTB detected. as compared with 98. There was no sig- for culture-confirmed tuberculosis was 92. All rights reserved. The user is provided with a printable test result. Two volumes of sample treatment reagent are added to each volume of sputum.4% in tum. For the detection of smear-negative. The sensitivity of a single direct MTB/RIF test HIV-negative patients (P = 0. with no significant vari. Appendix).24 by Jensen culture (Table 1 in the Supplementary chi-square test) (Table 2). Testing of multiple spec. No other uses without permission. direct MTB/ culture-positive cases and 90. All subsequent steps occur automatically. Next. Copyright © 2010 Massachusetts Medical Society. incubated at room temperature for 15 minutes. and shaken again. positive patients than did a single Löwenstein– ation in overall sensitivity across sites (P = 0. . 1010 n engl j med 363. the sensitivity of the MTB/ yield of a single assay performed directly on spu.

org on March 25.4) Diagnosis group at enrollment Suspicion of pulmonary tuberculosis — no. † In Azerbaijan.1) No tuberculosis — no.1) 217/304 (71. culture-negative tuberculosis 14/22 (63.and culture-positive tuberculosis — no.4) 90/230 (39.4) Copyright © 2010 Massachusetts Medical Society.2) * HIV denotes human immunodeficiency virus.6) 616/1730 (35. 2013.3) 30/346 (8.8) Smear-negative and culture-positive tuberculosis — 12/341 (3.3) 159/310 (51.2) 49/346 (14.2) 2/33 (6. No other uses without permission. (%) 75/316 (23./total no.0) 149/353 (42. (%) 22/341 (6. (%) 293/341 (85. For personal use only. 1011 ./total no.5) 69/353 (19.1) 9/102 (8.5) 15/346 (4.5)† 0/31 0/33 0/80 28/268 (10.0) 3/31 (9.7) 96/380 (25. (%) 102/341 (29.3) 115/268 (42.5) 52/380 (13.1) 12/380 (3.6) nies (LJ) Smear-positive. (%) 199/341 (58. (%) Distribution in final diagnostic category Smear.9) 31/380 (8.4) 4/91 (4. as compared with less than 2% for other sites./total no.11  nejm.1) 51/102 (50./total no.9) 1/31 (3.4) 80/353 (22.5) 23/268 (8.7) 272/346 (78.7) 15/346 (4.7) 21/380 (5.1) 73/353 (20.3) 567/1730 (32.7) 9/193 (4.1) 7/33 (21.9) 131/353 (37. Durban.2) 6/310 (1.2) 10/80 (12.1) 2/80 (2.3) 26/310 (8.1) 186/313 (59.9) 4/31 (12.5) 102/353 (28. Single culture positive with >28 days (MGIT) or <20 colo.8) 18/31 (58./total no.6) 148/344 (43. and MGIT mycobacteria growth indicator tube. Baku.5) 0/33 1/80 (1./total no.8) 59/346 (17.4) Suspicion of multidrug-resistant tuberculosis not receiving 41/341 (12.8) 189/380 (49.8) 268/1730 (15./total no./total no.4) 445/1730 (25. 0/22 0/102 3/31 (9. Demographic and Clinical Characteristics of the Patients.4) 1010/1730 (58.9) 12/33 (36.9) 105/1730 (6. there was a delay of up to 8 days between sputum collection and processing because of air transport to the laboratory in Germany.9%.4) 174/1730 (10. (%) Suspicion of multidrug-resistant tuberculosis receiving 7/341 (2.3) 49/80 (61. Mumbai.9) 70/353 (19.2) 136/380 (35.7) 219/346 (63. and Distribution in Final Diagnostic Category. which resulted in an overall culture con- tamination rate of 13. Table 1. LJ Löwenstein–Jensen.1) 223/380 (58.6) The New England Journal of Medicine Indeterminate — no. All rights reserved. Cape Town.5) 7/268 (2.6) Nontuberculous mycobacteria only 2/22 (9.9) therapy — no.5) 80/310 (25.3) 10/268 (3.9) 2/31 (6.2) 671/1447 (46.6) 5/102 (4. Peru Azerbaijan South Africa South Africa India All Patients Variable (N = 341) (N = 353) (N = 380) (N = 346) (N = 310) (N = 1730) Demographic or clinical characteristic Median age (range) — yr 31 (18–79) 37 (20–69) 36 (18–80) 32 (18–68) 30 (17–88) 34 (17–88) Female sex — no.2) History of tuberculosis — no. (%) 6/341 (1.4) 392/976 (40.8) 32/353 (9.9) tance receiving therapy Detection of Tuberculosis and Rifampin Resistance Contamination of ≥3 of 4 cultures 0/22 28/102 (27.2) 33/346 (9.3) 0/251 119/349 (34.7) 159/209 (76. Death or loss to follow-up 2/22 (9. (%) 138/319 (43. Diagnosis Group at Enrollment.1) 60/310 (19.6) 91/310 (29.5) n engl j med 363.3) 275/1730 (15./total no.1) no.1) 14/80 (17.3) 36/310 (11.* Lima.7) Downloaded from nejm.7) 137/251 (54.7) 162/310 (52.1) 4/102 (3. 2010 Culture-negative disease with suspected multidrug resis.4) 4/80 (5.5) 46/268 (17.0) 138/306 (45.1) 533/1462 (36.6) Discrepant phenotypic drug-susceptibility testing 2/22 (9.7) 10/33 (30.7) 2/33 (6.1) 5/102 (4. 2/22 (9.7) therapy — no./total no. (%) Clinical tuberculosis — no.0) 39/268 (14. (%) 3/179 (  september 9.5) HIV infection — no.1) 173/230 (75./total no.

of MTB/RIF tests 3 Samples (2 pellet and 1 direct) Correct — no.4–98.4–99.* Site and No.4–95.1) 199/199 (100) 10/12 (83. The n e w e ng l a n d j o u r na l of m e dic i n e Table 2.3% had positive results on the MTB/RIF test (data not shown). see Table 3 in the Supplementary Appendix.5) 604/609 (99.3) 95% CI 85.2–95.4) 162/162 (100. (%) 43/45 (95.0) 1127/1134 (99.0 65. At sites per.0) 47/52 (90.4–98.7–100. and  september 9.7 98. (%) 185/188 (98./total no./total no.5–99.6–98.0 71.0) 23/26 (88.7 Mumbai. For personal use only.0 84. According to the Number of Tests per Patient.1 94.6) 95% CI 94. P<0.1–96.7 98. .6 95.0 62. the sensitivity of the MTB/RIF test per. Overall Sensitivity and Specificity of the MTB/RIF Test.1–100.) Of 105 patients with culture-negative samples who were treated for tuberculosis on the ba- sis of clinical symptoms.5 Durban.1) 95% CI 92.8 96.6–99.0–100.1% for three MTB/RIF tests.9 90. Copyright © 2010 Massachusetts Medical Society.6) 566/567 (99. (%) 723/741 (97.8–99.9 96.7 79.2) 124/171 (72. (%)† 1423/1482 (96.6 * Site-specific performance is shown for three MTB/RIF test results per patient (two pellet samples plus one direct sam- ple).0) 13/15 (86.4) 95% CI 91. (%) 209/211 (99.5 No./total no. P = 0. as Compared with Three Smears and Four Cultures.6% vs.2–98.1–96.2) 604/616 (98.5) 35/36 ( on March 25. but no further analysis was done during this study. (%) 144/149 (96.6) 30/30 (100./total no.0 84.7–89. The specificity of the MTB/RIF test did not differ forming alternative nucleic acid–amplification significantly from that of Amplicor or Probetec testing. formed directly on sputum was higher than that of Amplicor (94. The second observation is a combination of the second sputum sample (pellet) and the third sputum sample (direct).8 95.96).8 79.0 85.2) 95% CI 90. The first observation is a combina- tion of the first sputum sample (pellet) and third sputum sample (direct).7) 213/219 (97.6) 80/80 (100.4–78. Azerbaijan Correct — no. (Table 2 in the Supplementary Appendix).96) per./total no.4–99.4) 186/189 (98. 86.5 99.9–93.8) 68/70 (97./total no.1–99. South Africa Correct — no.9 2 Samples (1 pellet and 1 direct) Correct — no.0–96./total no.01) and similar Detection of Multidrug Resistance to that of ProbeTec (83. All rights reserved.8%.3 94.4–100.1 98.2 Cape Town. 29.2–98.8–99.9%.9) 95/96 (99. 83.8 88.8) 157/174 (90.2) 551/561 (98. (%) 675/732 (92. 2013. of Tests Sensitivity Specificity Smear-Positive Smear-Negative All Culture-Positive and Culture-Positive and Culture-Positive No Tuberculosis Site Lima. 2010 The New England Journal of Medicine Downloaded from nejm. South Africa Correct — no. MTB/RIF test for the detection of rifampin and 1012 n engl j med 363. No other uses without permission.3–99.6–99.11  nejm.6–97.0 Baku.4–100.2–98. (%) 142/148 (95. The sensitivity of the test did not differ significantly between patients who were suspected of having pulmonary tu- berculosis and those suspected of having multidrug-resistant tuberculosis (P = 0.2 1 Sample (direct) Correct — no.3 96. India Correct — no.0–93.5 99.2 97. Table 3 shows the sensitivity and specificity of the formed on extracted DNA from sputum pellets. Peru Correct — no. The calculation of the confidence interval (CI) accounts for within-patient correlation and the use of the third sputum sample two times. (For details./total no.2) 95% CI 95. † The denominator for patients with two tests includes two observations per patient.1) 95% CI 96.0 55.2–99.1) 1215/1232 (98.3) 102/102 (100) 95% CI 96.0) 64/69 (92.4) 296/348 (85.6–100.7% vs.

51 had positive results on the MTB/RIF RIF test correctly detected rifampin resistance in test. which had been reported as resistant on phenotypic drug-susceptibility testing.0) 95% CI — % 94.9) 90/94 (95.0) 38/38 (100. negative tuberculosis who had suspected multi- tified 3 patients with mixed infection containing drug resistance and were receiving tuberculosis wild-type and mutant strains in the same culture. In a subgroup of 115 patients with culture- resistant on drug-susceptibility testing. All rights reserved.1) 209/211 (99. Of 723 culture-positive samples.1% sensitivity) and in all observed that all 8 patients were later started on 506 patients with rifampin susceptibility (100% second-line therapy for treatment failure by physi- specificity). 2013.1) 90/90 (100.5) 197/199 (99.6–99.0) 190/193 (98. none of 8 randomly representative of the global situation: 16 different selected patients from the same cohort with posi- mutations were identified. but a limited number. our data provide a first n engl j med 363. Copyright © 2010 Massachusetts Medical on March 25.0) 38/38 (100. determined the presence of a samples (55.9 96.0) 15/15 (100. (%) 119/121 (98. in samples from smear- phenotypic testing. negative.0) Mumbai.0) Durban.2) 62/62 (100. culture-positive patients.4–99. (%) 3/3 (100.3) 61/64 (95.9 96. No other uses without permission.2%).6) 505/515 (98. (%) 15/16 (93.3) 121/122 (99. not received treatment./total  september 9. Peru — no. According to Site. rently recommends that the MTB/RIF test be used pared the performance of the direct Genotype for patients with suspected tuberculosis who have MTBDRplus assay with that of the MTB/RIF test. culosis treatment.0) 126/126 (100./total no./total no. treatment (and who were excluded from the main Taking sequencing results into account.5–98. and rifampin resistance was detected in 8.7 99. 526 and 531. sequencing confirmed the phenotypic result: rpoB mutation 516 GTC was detected in 1.0) Cape Town.0) Total for rifampin resistance Correct — no. (%) 200/205 (97./total no. Although the manufacturer cur- Using the South African samples./total no. multidrug resistance (resistance to both rifampin In smear-positive sputum samples. During blinded sequencing of 15 discrepant samples. (%) 195/200 (97. Sensitivity and Specificity of the MTB/RIF Test for the Detection of Rifampin and Multidrug Resistance.0) 190/190 (100. India — no. In 2 samples. A wild-type allele was identified in 1 sample. (%) 47/49 (95. and iden. In comparison.7 * Multidrug resistance is defined as resistance to both rifampin and isoniazid.4–99. For 15 of 718 patients for whom assay showed a sensitivity equivalent to that of the results on the MTB/RIF test were discrepant on MTB/RIF test. For personal use only./total no. South Africa — no.7) 51/52 (98. Azerbaijan — no. Detection of Tuberculosis and Rifampin Resistance Table 3.0) Baku. (%) 16/16 (100.0) 3/3 (100. and 531 TTG in the other. for which the tance-associated rpoB mutations in nine strains that MTBDRplus assay is not indicated. 720 were analyzed for rifam­ pin resistance because results on the MTB/RIF test were indeterminate in 3 cases. South Africa — no.0 96.1) 506/506 (100. sequencing confirmed resis.3–98.11  nejm. However.2–100.4) 19/19 (100. the MTBDRplus were identified as rifampin-sensitive on drug-sus. we com. . the MTB/ analysis).0 Total for multidrug resistance Correct — no. the MTBDRplus and isoniazid).0) 95% CI — % 94.* Phenotypic Drug-Susceptibility Testing Site and Total Phenotypic Drug-Susceptibility Testing† and Discrepant Resolution by Sequencing† Sensitivity for Specificity for Sensitivity for Specificity for Rifampin Resistance Rifampin Resistance Rifampin Resistance Rifampin Resistance Lima. Mixed infections were identified in 3 samples and were excluded from the analysis after discrepant resolution. cians who were unaware of the results on MTB/ The rpoB mutations found in this study were RIF testing. accounted for rifampin resistance were given second-line tuber- almost all resistant strains. rpoB mutations were identified in 9 samples that were rifampin-sensitive on phenotypic drug-susceptibility testing. tive results on MTB/RIF testing that did not detect notably in codons 516. † This is the reference standard for the comparison with the MTB/RIF test.8) 126/126 (100. as Compared with Phenotypic Drug-Susceptibility Testing Alone and in Combination with Sequencing of Discrepant Cases. 2010 1013 The New England Journal of Medicine Downloaded from nejm. assay provided a false negative result in 37 of 67 ceptibility testing. wild-type allele in one strain deemed rifampin. We 209 of 211 patients (99. /total no.

especially in rural areas. Although FIND has ne- negative. the infra- ing more than 90% of patients with smear-negative structure and trained personnel required for such disease.11  nejm. is advantageous. No other uses without permission. suggesting that the findings often not available for at least 4  september 9. .org on March 25. even after culture conversion. samples (1. These in. plus its hands-on time of under 15 minutes in 381 of 6920 MGIT and Löwenstein–Jensen cul. ineffective tuberculosis detection and of-treatment use in low-income countries accurate. 2013.0%) with positive results had an inde. includ. the increased sensitivity of the MTB/RIF test MTB/RIF test automates DNA extraction. to a single nonprecise step that both liquefies and There are several reasons why the findings of inactivates sputum. However.2% (63 of 5190 tified as a challenge for implementation at periph- tests). qualitative questionnaires that were com- ing therapy.24.viable tubercle bacilli of 6 to 8 logs and eliminates proved care for patients with tuberculosis. culture-positive sputum samples with a gotiated concessionary pricing for public-sector very late cycle threshold (35 to 37 cycles) in the programs in low-income countries and is working MTB/RIF test. to achieve great simplicity of use. Large- repeat tests (92. The tion. The relative simplicity of the MTB/RIF than the overall culture-contamination rate (5. For personal use only. In view of the dramatically reduces its clinical utility. and detection inside a test cartridge that tuberculosis — at the two South African sites is never reopened.effective and highly sensitive detection of tubercu- ature and electricity supply will be more variable losis and drug resistance outside reference centers.the rise of multidrug resistance and extensively ly detected pulmonary tuberculosis and screened drug-resistant tuberculosis have led to calls for for rifampin resistance.and its unambiguous readout.8%). amplifi- — notably. with little chance of amplicon with 60 to 80% prevalence of HIV infection is contamination. and simplicity and safety of use could allow for cost- other point-of-treatment settings where temper. which results in a reduction in this study might not translate widely into im. the costs of say to analyze results for up to 40 cycles would have instruments and tests will still be considerably eliminated 19 of the 20 indeterminate results with. 1014 n engl j med 363. All rights reserved.higher than those for microscopy. among patients with smear-negative cation. A software change allowing the as. and recent study confirm that the MTB/RIF assay gen- it is not certain that our findings would be repli. The n e w e ng l a n d j o u r na l of m e dic i n e indication that the test also detects multidrug. In our study. Copyright © 2010 Massachusetts Medical Society.5%) costly to manufacture. is currently available in peripheral health care set- tings in many countries.erates no infectious aerosols. MTB/RIF test- Discussion ing could be less costly than implementation of culture and drug-susceptibility testing. No patient had indeterminate scale projects to show the feasibility and effect of results on all samples tested.23 Unfortunately. A total of 20 of 2072 MTB/RIF testing at such sites are under way. First. which is all that out affecting the specificity of the assay.ever.001). a rate that was lower microscopy).26 These features of cated in microscopy centers. How- resistant tuberculosis in patients who are receiv.25 The low sensitivity of smear microscopy for the diag.7%). Valid results were obtained in 129 of 139 eral laboratories. Data from a only reference facilities were used in the study.and training issues will be more relevant. Allowing for one repeat test.the disease is endemic. tures (P<0. the whereas the need for annual calibration was iden- indeterminate rate dropped to 1. Second. an assay that was designed for point. Specimen processing is simplified encouraging.complexity of standard nucleic acid–amplification nosis of tuberculosis in patients with HIV further lower the costs of testing. health posts. which determinate rifampin results all occurred in smear.MTB/RIF test uses sophisticated technology. Performance both for case detection and testing are not available except in a limited num- discrimination of rifampin resistance was similar ber of reference centers. the necessity for a biosafety cabinet. 2010 The New England Journal of Medicine Downloaded from nejm. Globally. the terminate result for rifampin resistance. pleted during the study suggested that users con- sidered 2 to 3 days a sufficient duration of training Indeterminate Rate for technicians without previous molecular experi- The MTB/RIF test was indeterminate in 192 of ence (as compared with 2 weeks for Ziehl–Neel­sen 5190 tests performed (3. This assay identified more dramatic expansion of culture capability and than 97% of all patients with culture-confirmed drug-susceptibility testing in countries in which tuberculosis who met the inclusion criteria. and results of testing are across diverse sites.tests prevents the expansion of this method. which are likely to be widely applicable.

STARD initiative: Standards for Reporting analysis for discrete and continuous out- ics. July 2006.2(1):e757. 19. Facing 11. Helb D. 10. 1968. Urdea M. Sivasubramani SK. ter E. and advanced biosafety. Biometrics 1986. FEMS Immunol Med Microbiol 2009. financing: WHO report. Bruns DE. Nature 2006. Waggoner RF. Copyright © 2010 Massachusetts Medical 2010 1015 The New England Journal of Medicine Downloaded from nejm. Bayona J. et al. conduct. Molecular detection.  september 9. Ching J. Clin Chem 2005. laboratory.48:229-37. J Infect Dis 2007. tibility testing sufficient? Int J Tuberc sis: an indicator of public-health negli. receive immediate notification when a journal article is released early To be notified when an article is released early on the Web and to receive the table of contents of the Journal by e-mail every Wednesday evening. Ranald Sutherland for their continuous technical support and ment) and grants from the National Institutes of Health (AI52523). Stop TB Partnership: the global plan A threshold value for the time delay to TB 14. Farmer P. Rieder HL. 5. Kra­ procedure manual for BACTEC MGIT 960 Copyright © 2010 Massachusetts Medical Society. Rapid (WHO/HTM/TB/2009. Nunn al. Clin Chem 2003.11  nejm. Becerra M. 15. Jones M. 2006. Atlanta: Centers for Disease 24. Bossuyt PM. McClatchy 25. J Clin berculosis in the HIV era. Public health to stop TB 2006–2015.) Lung Dis 2006. 4. Technology of the Xpert MTB/RIF Assay. mycobacteriology: a guide for the Level III Health Organization. et al.258. strategy. Mo. All rights reserved. Rüsch-Gerdes S. Kent PT. Hillemann D.369:2042-9. Xi L. Longitudinal data care in overlapping HIV and TB epidem. Piatek AS. sign up through our Web site at NEJM. timely completion of the study. Sanne I. PLoS One 2007. Siddiqi SH. et of binomial proportions. References 1.48:2495-501. infection or AIDS in resource-constrained PCR or reverse transcription-PCR clinical 21. gistics officer Nora Champouillon and data manager Pierrick velopment). Banada PP. Perkins MD. Yagui M.56: 17. Kubica GP. P. et al.42:121-30. et The dilemma of MDR-TB in the global Control. . 18. crease delays in diagnosis. Geneva: World diagnosis.39:4131-7.368:1554-6. 103-11. 2009. 6. Enarson D. 13. NJ: Beckton. Getahun H. Lancet 2007. MGIT gust 18 (Epub ahead of print). and rifampin resistance by use of on- the crisis: improving the diagnosis of tu. Gonnet for facilitating the project. 2013. Ruesch-Gerdes S. Application of the Capilia TB assay Blakemore R. JK. Helb D. Int J RIF assay and its applicability to point-of- tics in the developing world. tuberculosis. Geneva: World Health Organization. tecting drug resistance in Mycobacterium Microbiol 2010. aerosol infection risk by the Xpert MTB/ Requirements for high impact diagnos. Towards complete and accurate report. tection in Mycobacterium tuberculosis. gence. et al. Supported by FIND (for clinical evaluation and assay develop. Havlir DV. Cunningham J. Penny LA. comes. Liang KY. Evaluation of the analytical performance Nunn P. Approximate is settings: informing urgent policy chang. Perales MT. 26. No other uses without permission. Justine de Grandpré and Dr. Harrington M. mer FR. rapid. Palomino JC. um tuberculosis complex isolates. JAMA 2008.10:838-43. 16. Coull B. program conditions: is rapid drug suscep- 7. Story E. trol. Detection of Tuberculosis and Rifampin Resistance which would increase access to testing and de. Scaling up 8. demand.300:423-30. Rich. Geneva: World Health Organization. Pol AC. 52:119-26. Agresti A. Getahun H. 12. Tyagi S. Marras SA. El-Hajj HH.411. resistance in Mycobacterium tuberculosis Dickinson. 3. miology. Int J Tuberc Lung Dis 1998. Asencios L. eds. Opportunities and challenges for HIV ing of studies of diagnostic accuracy: the 22. Am Rev Respir Dis tionary on March 25. and quantitative biol 2010. Am Stat 1998. Int J Tuberc Lung Dis 2009. testing.9:1409-11. 23. for culture confirmation of Mycobacteri. et al. Tyagi S. Containment of bio- 9. Disclosure forms provided by the authors are available with the full text of this article at NEJM. The 7H11 medium for the cultivation tuberculosis culture services: a precau- identification and drug resistance de­ of mycobacteria. without the need to We thank the clinical and laboratory teams at the partner build large numbers of laboratories equipped for sites for their efforts in the implementation. Timely diagnosis of MDR-TB under era. Reitsma JB. Detection of rifampin TB System. et al. Warren RM. better than “exact” for interval estimation es. near-patient technology. Blakemore R. 63. Franklin Lakes.16:359. O’Brien R. XDR tuberculo. For personal use only. lecular beacon sequence analysis for de. Tuberc Lung Dis 2005. 13:799-800. Diagnosis of smear-negative pul. Lancet 2006. Story E. Nat Biotechnol 1998. Urbanczik R. J Clin Microbiol 2010 Au- 444:Suppl 1:73-9. 20. 196:Suppl 1:S15-S27. Global tuberculosis control — epide. Van Rie A. Zeger SL. Raja S.) J Clin Microbiol 2001. et al. care settings. of Diagnostic Accuracy. and the Bill and Melinda Gates Foundation (for assay de. Laboratory services in tuberculosis con. Olmsted SS. detection of Mycobacterium tuberculosis 2. J Clin Micro- monary tuberculosis in people with HIV for automated. Uys PW. Cohn ML.51:882-90. 49:1-6. (WHO/TB/98. van Helden PD.2:869-76. project management during assay development. 1985.98:295-6. and FIND lo- Cepheid. Alland D. n engl j med 363. in a single tube with molecular beacons.