Articles

Accuracy of the Xpert MTB/RIF test for the diagnosis of

pulmonary tuberculosis in children admitted to hospital in
Cape Town, South Africa: a descriptive study
Mark P Nicol, Lesley Workman, Washiefa Isaacs, Jacinta Munro, Faye Black, Brian Eley, Catharina C Boehme, Widaad Zemanay, Heather J Zar

Summary
Background WHO recommends that Xpert MTB/RIF replaces smear microscopy for initial diagnosis of suspected Published Online
HIV-associated tuberculosis or multidrug-resistant pulmonary tuberculosis, but no data exist for its use in children. July 18, 2011
DOI:10.1016/S1473-
We aimed to assess the accuracy of the test for the diagnosis of pulmonary tuberculosis in children in an area with
3099(11)70167-0
high tuberculosis and HIV prevalences.
See Online/Comment
DOI:10.1016/S1473-
Methods In this prospective, descriptive study, we enrolled children aged 15 years or younger who had been admitted 3099(11)70187-6
to one of two hospitals in Cape Town, South Africa, with suspected pulmonary tuberculosis between Feb 19, 2009, and Division of Medical
Nov 30, 2010. We compared the diagnostic accuracy of MTB/RIF and concentrated, uorescent acid-fast smear with a Microbiology and Institute
reference standard of liquid culture from two sequential induced sputum specimens (primary analysis). for Infectious Diseases and
Molecular Medicine
(Prof M P Nicol PhD,
Results 452 children (median age 194 months, IQR 111462) had at least one induced sputum specimen; 108 children W Zemanay PhD), Division
(24%) had HIV infection. 27 children (6%) had a positive smear result, 70 (16%) had a positive culture result, and 58 of Clinical Pharmacology,
Department of Medicine
(13%) had a positive MTB/RIF test result. With mycobacterial culture as the reference standard, MTB/RIF tests when
(L Workman MSc), and
done on two induced sputum samples detected twice as many cases (759%, 95% CI 645872) as did smear Department of Paediatrics and
microscopy (379%, 251508), detecting all of 22 smear-positive cases and 22 of 36 (611%, 444778) smear- Child Health (W Isaacs BSc,
negative cases. For smear-negative cases, the incremental increase in sensitivity from testing a second specimen was J Munro, F Black MBBCh,
B Eley FCPaed, Prof H J Zar PhD),
278% for MTB/RIF, compared with 138% for culture. The specicity of MTB/RIF was 988% (976999). MTB/
University of Cape Town, Cape
RIF results were available in median 1 day (IQR 04) compared with median 12 days (917) for culture (p<00001). Town, South Africa; National
Health Laboratory Service,
Interpretation MTB/RIF testing of two induced sputum specimens is warranted as the rst-line diagnostic test for Groote Schuur Hospital, Cape
Town, South Africa
children with suspected pulmonary tuberculosis.
(Prof M P Nicol, W Zemanay);
Red Cross War Memorial
Funding National Institutes of Health, the National Health Laboratory Service Research Trust, the Medical Research Childrens Hospital, Cape Town,
Council of South Africa, and Wellcome Trust. South Africa (W Isaacs, J Munro,
F Black, B Eley, Prof H J Zar); and
Foundation for Innovative
Introduction specimens is therefore needed, but culture can take weeks New Diagnostics (FIND),
Diagnosis of pulmonary tuberculosis in children has relied and is consequently unavailable to inform clinical Geneva, Switzerland
predominantly on clinical, radiological, and tuberculin decisions on initial treatment. A rapid diagnosis of (C C Boehme MD)

skin-test ndings.1 However, clinical diagnosis has low tuberculosis in children is desirable because delayed Correspondence to:
Prof Mark P Nicol,
specicity, radiological interpretation is subject to diagnosis is associated with poor outcome.7
5.28 Falmouth, Faculty of Health
interobserver variability, and the tuberculin skin test is a An urgent need therefore exists for a rapid, sensitive, Sciences, Anzio Road,
marker of exposure, not disease.13 Microbiological and specic test for tuberculosis and for identication of Observatory, Cape Town 7925,
conrmation with identication of drug resistance is drug-resistant disease in children. The performance of South Africa
[email protected]
increasingly important in the context of an emerging drug- Xpert MTB/RIF (Cepheid, Sunnyvale, CA, USA), an
resistant tuberculosis epidemic. Furthermore, conrmation integrated sample processing and nucleic acid amplication
is useful in children with HIV, in whom pill burden, drug test for detection of Mycobacterium tuberculosis and
interactions, and adherence issues make treatment of HIV resistance to rifampicin, was assessed in a large multicentre
and tuberculosis dicult. Use of repeated induced sputum study in adults with suspected tuberculosis.8 One MTB/
specimens in children is simple, well tolerated, and RIF test accurately detected the presence of tuberculosis in
eective for microbiological conrmation of pulmonary 982% of smear-positive and 725% of smear-negative
tuberculosis, even in infants.4,5 One induced sputum tuberculosis cases. Rifampicin resistance was detected
specimen provides a similar microbiological yield to three with a sensitivity of 991% and specicity of 100%. The test
gastric lavage specimens in children admitted to hospital results were available within 100 min of testing, much less
with pulmonary tuberculosis.4 time than for results from culture.
Smear microscopy is typically negative in children with WHO have endorsed the use of MTB/RIF as the initial
culture-conrmed tuberculosis, even when optimised diagnostic test in people suspected of having drug-
uorescence microscopy is used on concentrated resistant or HIV-associated tuberculosis,9 but no data are
specimens. 1,6 Mycobacterial culture of induced sputum available on its accuracy in children. We prospectively

www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0 1

 Articles

Consecutive children meeting the entry criteria were

1459 children screened enrolled, except when patients were discharged before
recruitment could be completed. Written, informed
967 excluded
consent for enrolment in the study was obtained from a
612 not eligible parent or legal guardian. The Research Ethics Committee
345 discharged of the University of Cape Town approved the study.
10 withdrew consent

Procedures
492 children enrolled A history and physical examination were done at
enrolment. Routine clinical investigations included chest
radiography, tuberculin skin test, and HIV testing in
40 excluded
33 without at least one valid
children whose HIV status was not known (HIV rapid test
induced sputum culture and in all children, followed by a conrmatory PCR for children
MTB/RIF result younger than 18 months or HIV ELISA for children aged
7 with positive culture from
another site 18 months or older). Children with HIV infection were
classied according to WHO clinical staging10 from stage 1
to stage 4. We recorded CD4 cell count and HIV viral load
452 children with one induced for children with HIV; children were classied according
sputum specimen
385 children with two induced to the US Centers for Disease Control and Prevention
sputum specimens immunological classication11 as category 1 to category 3
(no immune deciency to severe immune deciency). We
did CD4 cell counts with the panleucogating method and
viral load testing with Abbott RealTime HIV-1 assay
70 children with denite 216 with possible tuberculosis 166 without tuberculosis
tuberculosis 6 MTB/RIF positive 0 MTB/RIF positive (Abbott, Des Plaines, IL, USA). Two reviewers, masked to
52 MTB/RIF positive 0 smear positive 0 smear positive microbiological and other results, reported all chest
27 smear positive
radiographs according to a standardised format.
Children were followed up for the duration of their stay
Figure: Study prole
in hospital. The decision to start tuberculosis treatment
was at the discretion of the medical doctor caring for the
assessed use of MTB/RIF compared with culture of child. Follow-up ambulatory visits were done at 3 months
repeated induced sputum specimens for the diagnosis of for all children to assign a diagnostic category by
pulmonary tuberculosis in children. assessment of response to treatment or recovery in the
absence of tuberculosis treatment. Response to treatment
Methods or recovery was assessed at follow-up visits by recording
Participants symptoms, signs, and weight gain, and by repeating
We did a prospective study in the general paediatric chest radiograph at the completion of treatment. A study
wards of Red Cross War Memorial Childrens Hospital paediatrician who had access to all laboratory results
and Somerset Hospital, Cape Town, South Africa. made this assessment.
Children aged 15 years or younger were eligible for Sputum induction was done after a 23 h fast in a
enrolment if they had been admitted to hospital between dedicated sputum induction room by a trained research
Feb 19, 2009, and Nov 30, 2010, with pulmonary nurse, as previously described.4 A second induced
tuberculosis suspected on the basis of having a cough for sputum specimen was obtained, whenever possible, the
more than 14 days and one of the following: a household following day or a minimum of 4 h after the rst
contact infected with tuberculosis within the previous specimen. Baseline arterial pulse oximetry was done in
3 months, loss of weight or failure to gain weight in the all children; monitoring was done throughout the sputum
previous 3 months, a positive skin test to puried protein induction procedure and for 30 min thereafter.
derivative (PPD; 2TU, PPD RT23, Staten Serum Institute, Sputum specimens were processed within 2 h of
Denmark, Copenhagen), or a chest radiograph suggestive collection in an accredited routine diagnostic microbiology
of pulmonary tuberculosis. A positive skin test was laboratory by trained technicians who used standardised
dened as 5 mm or more of transverse induration in protocols. This laboratory also participated in the
children with HIV infection or 10 mm or more in children previously reported multicountry study of MTB/RIF.8
without HIV infection. Children were excluded if they After decontamination with N-acetyl-L-cysteine and
had received more than 72 h of tuberculosis treatment or sodium hydroxide (10% nal concentration), centrifuged
prophylaxis during their hospital admission, if they were sputum deposits were resuspended in 15 mL of
not resident in Cape Town and could not be followed up, phosphate buer. A drop of sediment was used for
if informed consent was not obtainable, or if an induced uorescent acid-fast smear microscopy. For MTB/RIF
sputum specimen could not be obtained. testing, 14 mL of MTB/RIF sample reagent was added to

2 www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0

 Articles

07 mL of the resuspended sputum pellet and

All Denite Possible Not tuberculosis
subsequently processed as previously reported.8 (n=452) (n=70) (n=216) (n=166)
Automated liquid culture (mycobacterial growth indicator,
Median (IQR) age (months) 194 237 176 183
BACTEC MGIT, Becton Dickinson Microbiology (111 to 462) (152 to 595) (106 to 406) (109 to 399)
Systems, Cockeysville, MD, USA) was done with 05mL Sex (male) 250 (55%) 39 (56%) 116 (54%) 95 (57%)
of the resuspended pellet. Cultures were incubated for HIV infection 108 (24%) 17 (24%) 55 (26%) 36 (22%)
6 weeks if negative. Positive cultures were identied by WHO clinical staging
acid-fast staining followed by MTBDRplus testing (Hain Stage 1 15 (14%) 2 (13%) 7 (47%) 6 (40%)
Lifesciences, Hehren, Germany)12 to conrm the presence Stage 2 43 (40%) 7 (16%) 20 (47%) 16 (37%)
of M tuberculosis and to test for resistance to rifampicin
Stage 3 27 (25%) 4 (15%) 13 (48%) 10 (37%)
and isoniazid. When clinically indicated, hospital sta
Stage 4 23 (21%) 4 (17%) 15 (65%) 4 (17%)
collected specimens from additional disease siteseg, in
HIV CDC immune suppression
children with suspected extrapulmonary and pulmonary
None 11 (10%) 2 (18%) 7 (64%) 2 (18%)
tuberculosis. Because culture results were not available
Moderate 34 (31%) 3 (9%) 16 (47%) 15 (44%)
at the time of MTB/RIF testing, sta doing and recording
Severe 54 (50%) 9 (17%) 28 (52%) 17 (32%)
MTB/RIF tests were unaware of culture results.
Unknown 9 (8%) 3 (33%) 4 (44%) 2 (22%)
If MTB/RIF or MTBDRplus tests identied the
History of tuberculosis 51 (11%) 7 (10%) 23 (11%) 21 (13%)
presence of rifampicin resistance, the corresponding
Radiological changes suggestive of 274 (64%) 44 (68%) 139 (68%) 91 (57%)
cultured isolate also underwent testing for phenotypic tuberculosis
resistance to rifampicin and isoniazid by automated
Started on tuberculosis treatment 216 (48%) 69 (99%) 147 (68%) 0
liquid MGIT culture.
Median (IQR) height for age Z score 150 158 169 128
On the basis of clinical and microbiological (25 to 05) (278 to 068) (27 to 071) (21 to 02)
investigations children were classied as having denite Median (IQR) weight for 15 177 152 124
tuberculosis (induced sputum culture positive for age Z score (23 to 06) (286 to 089) (237 to 065) (216 to 043)
M tuberculosis), not tuberculosis (negative tuberculosis Median (IQR) weight for height 056 093 039 039
cultures and documented resolution of symptoms and Z score (16 to 04) (229 to 028) (153 to 053) (124 to 032)
signs at 3 month follow-up visit in children who did not Malnutrition (weight for age 155 (343%) 31 (443%) 76 (352%) 48 (289%)
Z score <2)
receive treatment), or possible tuberculosis (all other
TST positive/TST result known (%)
children). Children with possible tuberculosis therefore
All children 128/372 (34%) 39/57 (68%) 78/176 (44%) 11/139 (8%)
included children not receiving treatment, who had no
documented recovery at follow-up, and all children placed HIV-infected 13/85 (15%) 3/11 (27%) 10/44 (23%) 0/30

on treatment in the absence of microbiological HIV-uninfected 115/287 (40%) 36/46 (78%) 68/132 (52%) 11/109 (10%)

conrmation. An automatic computed algorithm made Data are n (%) unless otherwise stated. CDC=US Centers for Disease Control and Prevention. TST=tuberculin skin test.
the assignment to diagnostic category.
Table 1: Baseline characteristics

Statistical analysis
The primary reference standard was a positive culture for
M tuberculosis from an induced sputum specimen. We Categorical data were summarised as proportions with
analysed patients separately with interpretable results 95% CIs. Statistical tests included two-sample test of
from at least one induced sputum specimen and those proportions, test, and Wilcoxon rank-sum test. All
with results from two induced sputum specimens. For statistical tests were two-sided at with an value of 005.
the primary analysis of the specicity of MTB/RIF we
included only those children with two interpretable MTB/ Results
RIF and culture results, because one negative culture We screened 1459 children, 492 of whom were enrolled
result is likely to miss a substantial proportion of culture- (gure). 452 children had MTB/RIF and culture test
conrmed cases.4 We excluded children with negative results from at least one induced sputum specimen and
induced sputum cultures who had a positive culture from 385 had test results from two specimens. 70 (16%) of
another (extrapulmonary) site from the analysis of 452 children were classied as having denite tuberculosis,
sensitivity and specicity, because of the diculty in 216 (48%) as having possible tuberculosis, and 166 (37%)
interpretation of MTB/RIF ndings in this group. The as not having tuberculosis (table 1). Most children with
sensitivity, specicity, and predictive values of the assays HIV had moderate or severe immune suppression, and
with 95% CIs were established. Data were analysed with many children had nutritional impairment (table 1).
Stata (version 10) and EpiInfo (version 6). Simple At the time of enrolment, 34 children had previously
descriptive statistics were used to characterise the study been given tuberculosis treatment for a median 1 day
population, normally distributed continuous data were no children received treatment for more than 3 days.
summarised by mean and 95% CI, and non-normally Almost all children with denite tuberculosis and most
distributed continuous data by median and IQR. children with possible tuberculosis were given

www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0 3

 Articles

Sensitivity Specicity Positive Negative Sensitivity (95% CI) for Sensitivity (95% CI) for
predictive predictive smear-positive denite smear-negative denite
value value tuberculosis tuberculosis
All children with complete results from at least one induced sputum specimen (n=452)
MTB/RIF
All 52/70, 743% (638848) 376/382, 984% (972997) 897% 954% 27/27, 100% (872100) 25/43, 581% (428735)
HIV-infected 17/17, 100% (805100) 91/91, 100% (960100) 100% 100% 10/10, 100% (692100) 7/7, 100% (590100)
HIV-uninfected 35/53, 660% (529792) 285/291, 979% (963996) 854% 941% 17/17, 100% (805100) 18/36, 50% (328672)
Smear microscopy 27/70, 386% (269503) 382/382, 100% (990100) 100% 899%
Children with complete results from two induced sputum specimens (n=385)*
MTB/RIF (rst induced sputum specimen) 34/58, 587% (456 711) 325/327, 994% (985100) 944% 931% 22/22, 100% (846100) 12/36, 333% (172495)
MTB/RIF (both induced sputum specimens)
All 44/58, 759% (645872) 323/327, 988% (976999) 917% 958% 22/22, 100% (846100) 22/36, 611% (444778)
HIV-infected 14/14, 100% (768100) 80/80, 100% (955100) 100% 100% 9/9, 100% (664100) 5/5, 100% (478100)
HIV-uninfected 30/44, 682% (539825) 241/245, 984% (96899.9) 882% 945% 13/13, 100% (753100) 17/31, 548% (363734)
Smear microscopy (two smears)
All 22/58, 379% (251508) 327/327, 100% (989100) 100% 901%

Data are number correct/number tested, % (95% CI). *Sensitivity and specicity calculated with results from both induced sputum cultures.

Table 2: Accuracy of MTB/RIF and smear for case detection with liquid culture as the reference standard

LPA/DST resistant LPA/DST sensitive LPA/DST inconclusive HIV (p=0042; table 2). We recorded more smear-positive
tuberculosis cases in children with HIV than in children
MTB/RIF resistant 2 0 0
without HIV, although the dierence was not signicant
MTB/RIF sensitive 0 70 1
(p=0072; table 2). The specicity of MTB/RIF in children
MTB/RIF indeterminate 1 4 0
with two induced sputum culture results available was
LPA/DST=culture-based drug susceptibility testing (line probe assay). 988%, with only four of 327 cases not detected (table 2).
Two additional children had a positive MTB/RIF test with
Table 3: Concordance between MTB/RIF and culture-based drug susceptibility testing for identication
of resistance to rifampicin (per-sample analysis) only one negative induced sputum culture result
available. All of these six children were classied as
having possible tuberculosis and had a documented
tuberculosis treatment on enrolment (table 1), and only response to treatment at 3 month follow-up, and probably
one childwho had possible tuberculosisdid not had tuberculosis. None of the 166 children in the not
improve on follow-up. tuberculosis group had a positive MTB/RIF result
We recorded at least one positive MTB/RIF test in (specicity 100%, 95% CI 978100). The specicity of
743% of children with denite tuberculosis (table 2), smear microscopy in this group was also 100%.
28% of children with possible tuberculosis, and no When results for rifampicin susceptibility testing were
children who did not have tuberculosis. MTB/RIF interpretable from both line probe assay and MTB/RIF,
detected all 27 smear-positive denite cases and 25 of MTB/RIF correctly identied all 70 rifampicin- susceptible-
43 smear-negative denite cases (table 2). cases and two rifampicin-resistant cases on a per-sample
When including only the 385 children with both culture analysis (table 3). However, we recorded one case of
and MTB/RIF results from two induced sputum rifampicin-resistant tuberculosis and four cases of
specimens, two MTB/RIF tests detected twice as many rifampicin-sensitive tuberculosis (by line-probe assay and
denite cases as did smear microscopy, including all conrmatory culture-based testing), which were reported
smear-positive cases but only 611% of smear-negative as indeterminate by MTB/RIF. Whenever rifampicin
cases (table 2). The rst MTB/RIF test detected all smear- resistance was identied by line-probe assay, resistance
positive cases but only a third of smear-negative cases; was conrmed by phenotypic susceptibility testing.
the second test increased the sensitivity of MTB/RIF for Fewer MTB/RIF tests were recorded as failures or
smear negative tuberculosis by 278%. By comparison invalid (one of 867, 0001%) than cultures were recorded
with MTB/RIF tests, a second test increased the sensitivity as contaminated (19 of 867, 22%; p<00001). MTB/RIF
of culture by 138% (eight of 58 cases). provided faster results, within a median of 1 day (IQR 04)
The proportion of children diagnosed with denite compared with 12 days (IQR 917) for culture.
tuberculosis was much the same in HIV-infected
(17 of 108, 157%) and uninfected children (53 of 344, Discussion
154%). The sensitivity of MTB/RIF was higher in Two MTB/RIF tests on induced sputum specimens
children with HIV than in children who did not have detected three-quarters of culture-conrmed tuberculosis

4 www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0

 Articles

in young children admitted to hospital with suspected

pulmonary tuberculosis, with very high specicity. MTB/ Panel: Research in context
RIF detected all smear-positive cases, but had a lower Systematic review
sensitivity in smear-negative cases, with two tests detecting We searched PubMed for studies about the Xpert MTB/RIF test published in English up to
about three-fths of cases. The yield of MTB/RIF was June 14, 2011, with the search terms Xpert or MTB/RIF and tuberculosis. We did not
twice that of smear microscopy. This test is widely identify any systematic reviews. We identied eight studies that assessed the use of MTB/RIF
anticipated to replace smear microscopy in resource-poor for detection of tuberculosis in respiratory specimens, all of which were in adult patients.8,14,1621
settings where HIV co-infection or drug-resistant
tuberculosis are common,9 and our results suggest that its Interpretation
use is a major improvement over use of smear microscopy. Most studies, apart from a large multicentre study8 and a multicentre demonstration study,14
Although time to detection was not a primary outcome for were small, with many including testing of archived samples. The sensitivity of MTB/RIF for
this study, MTB/RIF results were available within 1 day, detection of smear-positive tuberculosis varied between 95% and 100%, with most studies
which was substantially faster than for culture results. reporting sensitivity of 99100%. Sensitivity for smear-negative tuberculosis varied
The sensitivity for smear-negative disease was lower substantially, between 47% and 77%. In studies with adequate numbers of rifampicin-
than that previously reported in adults with suspected resistant samples, sensitivity for detection of rifampicin resistance varied between 94% and
tuberculosis of 85% for two MTB/RIF tests.8 The 100%, and specicity varied between 98% and 100%. Our study adds to these ndings by
incremental increase in sensitivity of a second MTB/RIF documenting the accuracy of MTB/RIF on induced sputum specimens for the detection of
test for smear-negative tuberculosis was substantial. culture-conrmed tuberculosis in children, a previously unreported population. We showed
WHO recommends one MTB/RIF test for adults with that the sensitivity of MTB/RIF for smear-negative tuberculosis is lower in children that it is in
suspected tuberculosis,9 but our ndings suggest that a adults, but is twice as sensitive compared with smear microscopy in children. We also showed
second test should be recommended in children who that the incremental benet in testing a second induced sputum specimen is substantial,
have a negative rst test. The benet of improved suggesting that, in children, a second specimen should be tested to optimise sensitivity.
sensitivity of a second test would need to be balanced
against the increased costs associated with repeat testing.
Although MTB/RIF was more sensitive than smear Sputum induction was not regarded as feasible for
microscopy, almost a quarter of children with culture- diagnosis in young children until about 6 years ago.
conrmed tuberculosis were negative on MTB/RIF testing; Several studies have now shown the ecacy and safety of
this proportion was even higher in smear-negative, culture- sputum induction in infants and young children,
positive children. Although a positive MTB/RIF test is particularly in those who have been admitted to hospital.5,14,15
helpful, a negative test should therefore be interpreted in However, data on the use of sputum induction in primary
the context of the childs clinical and radiological ndings. care facilities, where the major burden of childhood
Culture remains an important diagnostic method. tuberculosis occurs, are scarce (panel). If MTB/RIF testing
Most children were treated for tuberculosis on the basis is to be implemented at or close to the point of care, then
of clinical rather than microbiological ndings. In view of the capacity for sputum induction in children at health
the fact that an ideal test for tuberculosis in children is facilities will need to be scaled up. More widespread use of
likely to have better sensitivity than culture, culture might sputum induction in children poses operational
be an imperfect reference standard. The absence of a highly challenges, including training of sta, operator time, and
sensitive reference standard makes assessment of this and the need for precautions to prevent transmission. For
other, potentially more sensitive, future technologies ambulatory children, two induced sputum procedures
challenging. The development of better diagnostic methods necessitate a second visit or an extended stay at a health
for childhood tuberculosis remains a major priority.13 facility. In our study, MTB/RIF testing was done at a large
In this study, MTB/RIF detected all cases of denite diagnostic laboratory in an academic centre; however, the
tuberculosis in children with HIV. However, because the test is robust and can be easily and competently done at
number of children with HIV and culture-proven disease microscopy centres.22
was small, further study is needed to conrm whether MTB/RIF is designed to detect not only the presence of
sensitivity is increased in these patients. tuberculosis, but also RIF resistance. This study was
The specicity of MTB/RIF was high with only six primarily aimed at assessment of the use of MTB/RIF for
children having a positive MTB/RIF test and at least one tuberculosis detection. Because few rifampicin-resistant
negative induced sputum culture. Interpretation of the cases were detected in this study, we are unable to draw
results for these six children is dicult, because of the conclusions about the ability of MTB/RIF to detect
poor sensitivity of culture for the diagnosis of childhood resistance. Further limitations of this study include the few
tuberculosis.1 However, these children probably had children with HIV infection and culture-conrmed disease,
pulmonary tuberculosis because they all had a good and the need to split sputum sediment between culture
clinical response to treatment at follow-up visits. The and MTB/RIF testing. MTB/RIF is designed for use directly
true specicity of MTB/RIF might be even higher, on sputum specimens, rather than on sputum pellet.
because none of the 166 children in whom tuberculosis However a study in adults showed equivalent performance
was excluded had a positive MTB/RIF test. for both specimen types.8 For this study we chose to split

www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0 5

 Articles

the sputum pellet for tests to obtain a direct comparison of 3 Hesseling AC, Schaaf HS, Gie RP, Starke JR, Beyers N. A critical
culture and MTB/RIF on the same specimen. In view of review of diagnostic approaches used in the diagnosis of childhood
tuberculosis. Int J Tuberc Lung Dis 2002; 6: 103845.
the paucibacillary nature of childhood tuberculosis, had we 4 Zar HJ, Hanslo D, Apolles P, Swingler G, Hussey G. Induced
processed one specimen by MTB/RIF and the other by sputum versus gastric lavage for microbiological conrmation
culture, we would have had many discrepant results, which of pulmonary tuberculosis in infants and young children:
a prospective study. Lancet 2005; 365: 13034.
would have been dicult to interpret. Because induced 5 Zar HJ, Tannenbaum E, Apolles P, Roux P, Hanslo D, Hussey G.
sputum specimens are typically very low volume, and Sputum induction for the diagnosis of pulmonary tuberculosis in
M tuberculosis bacilli are probably lost during the specimen infants and young children in an urban setting in South Africa.
Arch Dis Child 2000; 82: 30508.
decontamination process, direct testing of induced sputum 6 Marais BJ, Pai M. Recent advances in the diagnosis of childhood
specimens with MTB/RIF (when the whole specimen tuberculosis. Arch Dis Child 2007; 92: 44652.
volume would be used) might have improved sensitivity 7 Viani RM, Lopez G, Chacon-Cruz E, Hubbard P, Spector SA.
compared with testing of the pellet. Poor outcome is associated with delayed tuberculosis diagnosis
in HIV-infected children in Baja California, Mexico.
This study enrolled children who were admitted to a Int J Tuberc Lung Dis 2008; 12: 41116.
secondary or tertiary care facility, who could have had 8 Boehme CC, Nabeta P, Hillemann D, et al. Rapid molecular
more severe disease than those seen in primary care detection of tuberculosis and rifampin resistance. N Engl J Med
2010; 363: 100515.
settings. However, these hospitals are major referral 9 WHO. Tuberculosis diagnostics: Automated DNA test. http://www.
centres for children across the Greater Cape Town region. who.int/tb/features_archive/xpert_factsheet.pdf (accessed
This urban area has one of the highest rates of tuberculosis April 26, 2011).
10 WHO. WHO case denitions of HIV for surveillance and revised
in South Africa, but has good access to health services.23 clinical staging and immunological classication of HIV-related
The performance of MTB/RIF might dier in children disease in adults and children, 2006. http://www.who.int/hiv/pub/
with less severe illness or in other settings where the guidelines/HIVstaging150307.pdf (accessed July 5, 2011).
11 Caldwell MB, Oxtoby MJ, Simonds RJ, Lindegren ML, Rogers MF.
spectrum of illness is not the same because of diering 1994 revised classication system for human immunodeciency
access to health care; further studies are needed. Further virus infection in children less than 13 years of age. Centers for
work is also needed to assess MTB/RIF with other, easily Disease Control and Prevention. MMWR 1994; 43: 110.
obtained specimen types, such as urine and stool. 12 Barnard M, Albert H, Coetzee G, OBrien R, Bosman ME. Rapid
molecular screening for multidrug-resistant tuberculosis in a
MTB/RIF is a reliable test for rapid diagnosis of high-volume public health laboratory in South Africa.
tuberculosis in children when used on induced sputum Am J Respir Crit Care Med 2008; 177: 78792.
specimens. Testing two specimens substantially increases 13 Newton SM, Brent AJ, Anderson S, Whittaker E, Kampmann B.
Paediatric tuberculosis. Lancet Infect Dis 2008; 8: 498510.
diagnostic yield for smear-negative tuberculosis. To 14 Shata AM, Coulter JB, Parry CM, Chingani G, Broadhead RL,
maximise the eect of this technology and benet child Hart CA. Sputum induction for the diagnosis of tuberculosis.
health, increased capacity for sputum induction in Arch Dis Child 1996; 74: 53537.
15 Wiersma HE, Van Aalderen WM, Hoekstra MO. Sputum induction
children at health-care facilities is needed. for the diagnosis of pulmonary tuberculosis. Arch Dis Child 2000;
Contributors 83: 276.
MPN and HJZ had the idea for and supervised the study, obtained 16 Armand S, Vanhuls P, Delcroix G, Courcol R, Lemaitre N. Comparison
funding, analysed the results and primarily drafted the Article. WZ of the Xpert MTB/RIF test with an IS6110-TaqMan real-time PCR assay
coordinated the microbiological testing. LW and JM were responsible for for direct detection of Mycobacterium tuberculosis in respiratory and
data management and analysis. WI and FB enrolled patients and did nonrespiratory specimens. J Clin Microbiol 2011; 49: 177276.
sputum induction and data collection. BE supervised clinical sta and 17 Friedrich SO, Venter A, Kayigire XA, Dawson R, Donald PR,
contributed to study data. CCB provided expert advice regarding Diacon AH. Xpert MTB/RIF and Genotype MTBDRplus for patient
MTB/RIF testing and protocols. All authors reviewed the nal paper. selection for a tuberculosis clinical trial (JCM00138-11). J Clin Microbiol
2011; published online June 8. DOI:10.1128/JCM.00138-11.
Conicts of interest 18 Helb D, Jones M, Story E, et al. Rapid detection of
MPN has received funding from the Foundation for Innovative New Mycobacterium tuberculosis and rifampin resistance by use
Diagnostics (FIND, Geneva, Switzerland) to assess the performance and of on-demand, near-patient technology. J Clin Microbiol 2010;
eect of MTB/RIF. CCB is employed by FIND. FIND is a non-prot 48: 22937.
organisation that collaborates with industry partners, including Cepheid 19 Marlowe EM, Novak-Weekley SM, Cumpio J, et al. Evaluation of the
(the manufacturers of the Xpert MTB/RIF test), on the development, Cepheid Xpert MTB/RIF assay for direct detection of
assessment, and demonstration of new diagnostic tests. All other Mycobacterium tuberculosis complex in respiratory specimens.
authors declare that they have no conicts of interest. J Clin Microbiol 2011; 49: 162123.
20 Moure R, Munoz L, Torres M, Santin M, Martin R, Alcaide F. Rapid
Acknowledgments detection of Mycobacterium tuberculosis complex and rifampin
This study was funded by the National Institutes of Health, USA resistance in smear-negative clinical samples by use of an
(1R01HD058971-01), the National Health Laboratory Service Research integrated real-time PCR method. J Clin Microbiol 2011; 49: 113739.
Trust, the Medical Research Council of South Africa and the Wellcome 21 Theron G, Peter J, van Zyl-Smit R, et al. Evaluation of the Xpert(R)
Trust (085251/B/08/Z). We thank the National Health Laboratory Service MTB/RIF assay for the diagnosis of pulmonary tuberculosis
diagnostic microbiology at Groote Schuur Hospital, the children who in a high HIV prevalence setting. Am J Respir Crit Care Med 2011;
participated in the study, and the childrens carers. published online April 14. DOI:10.1164/rccm.201101-0056OC.
22 Boehme CC, Nicol MP, Nabeta P, et al. Feasibility, diagnostic
References accuracy, and eectiveness of decentralised use of the Xpert
1 Zar HJ, Connell TG, Nicol M. Diagnosis of pulmonary tuberculosis MTB/RIF test for diagnosis of tuberculosis and multidrug resistance:
in children: new advances. Expert Rev Anti Infect Ther 2010; 8: 27788. a multicentre implementation study. Lancet 2011; 377: 1495505.
2 Hatherill M, Hanslo M, Hawkridge T, et al. Structured approaches 23 Day C, Gray A. Health and related indicators. In: Ijumba P,
for the screening and diagnosis of childhood tuberculosis in a high Padarath A, eds. South African Health Report 2006. Durban: Health
prevalence region of South Africa. Bull World Health Organ 2010; Systems Trust, 2006: 369506.
88: 31220.

6 www.thelancet.com/infection Published online July 18, 2011 DOI:10.1016/S1473-3099(11)70167-0