Articles

Xpert MTB/RIF Ultra versus mycobacterial growth indicator

tube liquid culture for detection of Mycobacterium tuberculosis
in symptomatic adults: a diagnostic accuracy study
Yingda L Xie, Christie Eichberg, Nchimunya Hapeela, Elizabeth Nakabugo, Irene Anyango, Kiranjot Arora, Jeffrey E Korte, Ronald Odero,
Judi van Heerden, Widaad Zemanay, Samuel Kennedy, Pamela Nabeta, Mahmud Hanif, Camilla Rodrigues, Alena Skrahina, Wendy Stevens,
Reynaldo Dietze, Xin Liu, Jerrold J Ellner, David Alland, Moses L Joloba, Samuel G Schumacher, Kimberly D McCarthy, Lydia Nakiyingi, Susan E Dorman

Summary
Background Xpert MTB/RIF Ultra (Ultra) is an automated molecular test for the detection of Mycobacterium Lancet Microbe 2024
tuberculosis in sputum. We compared the sensitivity of Ultra to that of mycobacterial growth indicator tube (MGIT) Published Online
liquid culture, considered the most sensitive assay in routine clinical use. https://doi.org/10.1016/
S2666-5247(24)00001-6

Methods In this prospective, multicentre, cross-sectional diagnostic accuracy study, we used a non-inferiority design to Department of Medicine,
Rutgers New Jersey Medical
assess whether the sensitivity of a single Ultra test was non-inferior to that of a single liquid culture for detection of
School, Newark, NJ, USA
M tuberculosis in sputum. We enrolled adults (age ≥18 years) with pulmonary tuberculosis symptoms in 11 countries (Y L Xie MD, K Arora MPH,
and each adult provided three sputum specimens with a minimum volume of 2 mL over 2 days. Ultra was done J J Ellner MD, D Alland MD);
directly on sputum 1, and Ultra and MGIT liquid culture were done on resuspended pellet from sputum 2. Department of Medicine,
Results of MGIT and solid media cultures done on sputum 3 were considered the reference standard. The Medical University of South
Carolina, Charleston, SC, USA
pre-defined non-inferiority margin was 5⋅0%. (C Eichberg BS, J E Korte PhD,
S Kennedy MPH, S E Dorman MD);
Findings Between Feb 18, 2016, and Dec 4, 2019, we enrolled 2906 participants. 2600 (89%) participants were analysed, Division of Medical Microbiology
including 639 (25%) of 2600 who were positive for tuberculosis by the reference standard. Of the 2357 included in the and Institute for Infectious
Diseases and Molecular
non-inferiority analysis, 877 (37%) were HIV-positive and 984 (42%) were female. Sensitivity of Ultra performed
Medicine, University of Cape
directly on sputum 1 was non-inferior to that of sputum 2 MGIT culture (MGIT 91⋅1% vs Ultra 91⋅9%; difference Town, Cape Town, South Africa
–0⋅8 percentage points; 95% CI –2⋅8 to 1⋅1). Sensitivity of Ultra performed on sputum 2 pellet was also (N Hapeela MPhil,
non-inferior to that of sputum 2 MGIT (MGIT 91⋅1% vs Ultra 91⋅9%; difference –0⋅8 percentage points; –2⋅7 to 1⋅0). J van Heerden MSc,
W Zemanay PhD); Infectious
Diseases Institute, Makerere
Interpretation For the detection of M tuberculosis in sputum from adults with respiratory symptoms, there was no University, Kampala, Uganda
difference in sensitivity of a single Ultra test to that of a single MGIT culture. Highly sensitive, rapid molecular (E Nakabugo BSc,
approaches for M tuberculosis detection, combined with advances in genotypic methods for drug resistance detection, L Nakiyingi PhD); Kenya Medical
have potential to replace culture. Research Institute, Center for
Global Health Research, Kisumu,
Kenya (I Anyango BSN,
Funding US National Institute of Allergy and Infectious Diseases. R Odero BSc); FIND, Geneva,
Switzerland (P Nabeta MD,
Copyright © 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license S G Schumacher PhD); State TB
(http://creativecommons.org/licenses/by/4.0/). Training and Demonstration
Centre, New Delhi, India
(M Hanif PhD); PD Hinduja
Introduction have confirmed the high sensitivity of Ultra for the diagnosis Hospital and Medical Research
The Xpert MTB/RIF tests (Cepheid, Sunnyvale, CA) are of pulmonary tuberculosis in adults and children and for Centre, Mumbai, India
automated, integrated, cartridge-based molecular assays the diagnosis of extrapulmonary tuberculosis.7–9 WHO (C Rodrigues MD); National
intended for tuberculosis case detection and identification endorsed Ultra in 2017 and it has been rolled out in national Reference Laboratory,
Republican Scientific and
of rifampicin resistance.1–3 The original Xpert MTB/RIF tuberculosis programmes.10 Practical Centre for
assay (ie, Xpert) was endorsed by WHO in 2010.4 Xpert Mycobacterial culture using liquid media has been Pulmonology and Tuberculosis,
MTB/RIF Ultra (ie, Ultra) was designed to address several the clinical gold standard for the diagnosis of active tuber- Minsk, Belarus (A Skrahina MD);
limitations of the original test. Most notably, Ultra incor- culosis for about three decades. Nevertheless, mycobacterial Department of Molecular
Medicine and Hematology,
porates two different multicopy amplification targets to culture has important practical and technical limitations.
Faculty of Health Science, School
increase test sensitivity.5 Analytical laboratory studies Growth in culture requires days to weeks and substantial of Pathology, and the National
demonstrated approximately one log improvement in the laboratory infrastructure for test performance and biosafety. Priority Program of the National
lower limit of detection for Ultra compared with Xpert.5 Cultures are subject to overgrowth by contaminating Health Laboratory Service,
A multicentre clinical diagnostic accuracy study confirmed microorganisms and might yield no clinically useful infor- Johannesburg, South Africa
(W Stevens MBBCh);
that Ultra was more sensitive than Xpert overall as well as in mation for some specimens. Preparation of respiratory Universidade Federal do Espirito
a subgroup with paucibacillary pulmonary disease and a specimens for culture requires exacting processing steps Santo, Vitoria, Brazil
subgroup with HIV, although Ultra had lower specificity that can decrease viable mycobacteria, thereby reducing (R Dietze MD);
than Xpert.6 Other large clinical diagnostic accuracy studies sensitivity.11–13 There are also Mycobacterium tuberculosis

www.thelancet.com/microbe Vol ▪ ▪ 2024 1

 Articles

Henan Provincial Chest

Hospital, Zhengzhou, China Research in context
(X Liu MD); Mycobacteriology
Laboratory, Department of Evidence before this study growth indicator tube (MGIT) culture for the detection of
Microbiology, School of Liquid culture is considered the most sensitive routinely available M tuberculosis in sputum specimens. Using a culture-based
Biomedical Sciences, Makerere test for the detection of Mycobacterium tuberculosis in clinical reference standard, we made two comparisons—namely, Ultra
University, Kampala, Uganda
specimens. Xpert MTB/RIF Ultra (Ultra) has also been shown to versus MGIT done on the same processed resuspended sputum
(M L Joloba PhD); US Centers for
Disease Control and have high diagnostic sensitivity; however, almost all published pellet and Ultra done directly on one unprocessed sputum
Prevention, Kisumu, Kenya studies report the sensitivity of Ultra testing with a single sputum specimen versus MGIT done on the processed sputum pellet from a
(K D McCarthy MS) specimen using a microbiological reference standard comprising different sputum specimen. Ultra was as sensitive as MGIT in both
Correspondence to: liquid cultures of multiple sputum specimens. To ascertain whether of these comparisons.
Dr Susan E Dorman, Department head-to-head comparisons of the diagnostic sensitivity of a single
of Medicine, Medical University Implications of all the available evidence
liquid culture versus that of a single Ultra test had been done,
of South Carolina, Charleston, Our results showing that Ultra has a sensitivity that is similar to
SC 29425, USA we searched PubMed with the terms ((pulmonary tuberculosis or
that of MGIT are novel. This information will help policy makers
[email protected] pulmonary TB) AND Ultra AND (MGIT or liquid culture) AND
and tuberculosis programmes calculate the relative merits and
comparative sensitivity or head-to-head)) for publications
drawbacks of diagnostic testing strategies that incorporate these
between Jan 1, 2008, and Feb 8, 2023. We identified and reviewed
assays. Our findings support a transition away from culture-based
13 publications. No publication described head-to-head
tuberculosis diagnostic algorithms and towards molecular
comparisons of Ultra and liquid culture in adults who are
approaches for pulmonary tuberculosis case detection. The crucial
symptomatic for tuberculosis.
next step to this transition requires biological and technical
Added value of this study advances that yield comprehensive drug susceptibility information
To our knowledge, we present the first evaluation of the from molecular tests performed directly on sputum.
comparative diagnostic sensitivities of Ultra and mycobacterial

bacterial phenotypes that do not grow readily in routine Study participants were adults (age ≥18 years) presenting
culture media.14,15 with pulmonary tuberculosis symptoms of cough for at least
The advent of highly sensitive rapid molecular tests raises 2 weeks plus at least one additional tuberculosis sign or
the possibility of transforming the clinical diagnostic symptom (eg, fever, night sweats, weight loss) and indica-
process by eliminating culture. In this study, we sought tion for pulmonary tuberculosis diagnostic evaluation per
to test the hypothesis that, for pulmonary tuberculosis case national tuberculosis programme guidelines. Participants
detection, the sensitivity of a single sputum Ultra test formed a consecutive series (ie, all eligible participants were
is non-inferior to that of a single mycobacterial liquid culture offered enrolment). Individuals were excluded as ineligible
performed from sputum using the BACTEC Mycobacterial if they had received any tuberculosis antimicrobials within
Growth Indicator Tube (MGIT) automated system 6 months before enrolment or were enrolled into the
(BD Microbiology Systems, Sparks, MD, USA). multidrug resistance risk group of the previous cohort.6
Demographic information, medical history, and HIV
Methods status were recorded at enrolment. Classification of a par-
Study design and procedures ticipant as having had (versus not) previously treated
In this diagnostic accuracy study, we combined data from a tuberculosis was based on participant self-report. Partic-
cohort of newly enrolled, newly tested participants with data ipants were asked to provide three spontaneously expecto-
from a previously published cohort.6 For both cohorts, rated sputum specimens with a minimum volume of 2 mL
specimens and microbiological testing were prospectively per specimen over 2 days. The interval between each
See Online for appendix conducted according to written protocols (appendix p 5) and sputum collection was at least 1 h, with the third sputum
data were collected on study-specific forms. In the original collection targeted as a morning sample on the second day.
cohort, the primary study objective was to determine Ultra Tests were performed at on-site study laboratories according
diagnostic accuracy versus a microbiological reference to written procedures (figure 1).
standard comprising multiple sputum cultures. Partic- For sputum sample 1, smear microscopy was done using
ipants in the original cohort were enrolled at primary health auramine–rhodamine staining and visualised bacilli burden
centres and hospitals in eight countries (Belarus, Brazil, was graded as negative, scanty, 1+, 2+, or 3+ by the laboratory
China, Georgia, India, Kenya, South Africa, and Uganda). microscopist.16 Xpert and Ultra assays were done by adding
The current study extended enrolment at sites in Kenya, kit-provided sample reagent in a 2:1 dilution to the sputum
South Africa, and Uganda, using the same case detection sample and adding 2⋅0 mL of the resulting mixture to one
group eligibility criteria and specimen testing procedures as Xpert cartridge and one Ultra cartridge. Samples were ana-
reported previously.6 Per the protocol, extension of enrol- lysed using GeneXpert instruments with automated results
ment was done explicitly to assess for non-inferiority of readouts of M tuberculosis detected, M tuberculosis not
Ultra sensitivity versus MGIT. detected, invalid, error, or no result. The semiquantitative

2 www.thelancet.com/microbe Vol ▪ ▪ 2024

 Articles

Sputum 1 Sputum 2 Sputum 3

N-acetyl-L-cysteine and N-acetyl-L-cysteine and

sodium hydroxide processing sodium hydroxide processing

Smear Smear Smear

microscopy microscopy microscopy

Xpert-Sp1-raw index test MGIT-Sp2-pellet comparator test

Xpert MGIT liquid

MGIT liquid culture
G4 culture

Comparison Reference
Xpert-Sp1-raw index test A Löwenstein–Jensen-Sp2-pellet standard
index test
Xpert Löwenstein–Jensen
Ultra Löwenstein– culture
Comparison
Jensen
B
culture

Xpert-Sp2-pellet index test

Xpert
Ultra

Figure 1: Schema for sputum testing and comparison of diagnostic accuracy of Ultra vs MGIT liquid culture
The reference standard was the culture result from sputum 3; both MGIT and Löwenstein–Jensen culture results were used. The comparator test was the MGIT culture done on the
resuspended pellet from sputum 2 (MGIT-Sp2-pellet). For primary comparison A, the index test was the Ultra test done directly on unprocessed sputum 1 (Ultra-Sp1-raw). For
primary comparison B, the index test was the Ultra test done on the resuspended pellet from sputum 2 (Ultra-Sp2-pellet). MGIT=mycobacterial growth indicator tube liquid culture.

scale for Ultra results was trace, very low, low, medium, or The study was reviewed and approved by institutional
high. A result of detection (including trace) of M tuberculosis review or ethics committees at each site. Written informed
was considered positive regardless of semiquantitative results. consent was obtained from all participants.
For sputum specimens 2 and 3, each specimen was
digested with N-acetyl-L-cysteine and sodium hydroxide Statistical analysis
(final concentration 1⋅0–1⋅5%), which was concentrated The reference standard was the sputum 3 culture result,
using centrifugation, and the resulting pellet was resus- inclusive of results of MGIT and Löwenstein–Jensen
pended in phosphate-buffered saline (pH 6⋅8) to a final medium. If either or both of the reference tests were positive
volume of 2⋅0 mL.17 Smear microscopy was done using for M tuberculosis, then the reference standard was consid-
auramine–rhodamine staining (50 μL) of the pellet and ered positive for M tuberculosis. If both were negative for
graded as negative, scanty, 1+, 2+, or 3+. For cultures, 0⋅5 mL M tuberculosis, or if one was negative and the other was
of the resuspended pellet was inoculated into a MGIT liquid contaminated, then the reference standard was considered
culture vial that was incubated using the BACTEC negative for M tuberculosis (figure 1). The comparator test
960 instrument (BD Microbiology Systems) and 0⋅2 mL of was MGIT culture on the resuspended pellet from sputum 2
this was inoculated onto Löwenstein–Jensen medium (MGIT-Sp2-pellet). Two comparisons were prespecified.
(BD Microbiology Systems).18,19 Cultures that were positive For comparison A, the index test was the Ultra result from
for the growth of acid-fast bacilli underwent confirmation of unprocessed sputum 1 (Ultra-Sp1-raw). For comparison B,
M tuberculosis complex by MPT64/MPB64 antigen detec- the index test was the Ultra result from the resuspended
tion.18 Culture results were classified as contaminated, pellet from sputum 2 (Ultra-Sp2-pellet). We also determined
M tuberculosis detected, or no M tuberculosis detected. For the sensitivities of Xpert using unprocessed sputum 1
Ultra testing, 1⋅4 mL sample reagent was added to 0⋅7 mL of (Xpert-Sp1-raw) and Löwenstein–Jensen culture using
the resuspended pellet (2:1 dilution) and 2⋅0 mL of the resuspended pellet from sputum 2 (LJ-Sp2-pellet).
resulting mixture was added to one Ultra cartridge and Ultra and Xpert results of M tuberculosis detected and
tested as described. M tuberculosis not detected were classified as determinate.
Staff performing Ultra and Xpert assays were masked to the Results of invalid, error, and no result were classified as
results of the other study tests through specimen codes and non-determinate. Cultures with growth of M tuberculosis or
staffing assignments. Laboratory staff did not have access no growth of M tuberculosis were classified as determinate,
to participants’ clinical information. Data were captured and contaminated cultures were classified as non-
through dedicated password-protected data entry systems. determinate. A participant was classified as smear-positive

www.thelancet.com/microbe Vol ▪ ▪ 2024 3

 Articles

if any study smear was positive for acid-fast bacilli by smear

microscopy or as smear-negative if all study smears were 3489 potentially eligible participants
negative for acid-fast bacilli (smear scanty results were
considered positive). Sensitivity was defined as the propor-
583 participants excluded
tion of participants testing positive for M tuberculosis with 371 participants enrolled into
the reference standard who had also tested positive for prespecified drug resistance group
212 individuals declined to participate
M tuberculosis with the test of interest. Specificity was
defined as the proportion of participants testing negative for
M tuberculosis using the reference standard who also tested 2906 enrolled participants
negative for M tuberculosis using the test of interest.
Participants who provided fewer than three sputum
specimens or for whom the reference standard result was 306 participants excluded
242 with fewer than three sputum
non-determinate were excluded from analyses. CIs for specimens collected
simple proportions were calculated using the Wilson score 64 reference standard non-determinate
interval. due to contamination
The primary hypothesis was that the sensitivity of a single
Ultra test was non-inferior to that of a single MGIT culture. 2600 with a determinate reference standard
To assess non-inferiority, the upper limit of the 95% CI of 639 with a positive reference standard
the difference in sensitivity (expressed as comparator sen- 1961 with a negative reference standard

sitivity minus index test sensitivity) was compared with the

predefined non-inferiority margin of 5⋅0 percentage points. 243 participants excluded*
Non-inferiority was considered to be achieved if the upper 47 Sp1 ultra not done
4 Sp1 ultra result non-determinate
limit of the 95% CI of the sensitivity difference was no 26 Sp2 ultra not done
greater than the non-inferiority margin. The margin of non- 8 Sp2 ultra result non-determinate
inferiority was selected based on informal clinician input 167 Sp2 MGIT contaminated

with regard to the potential benefits of a rapid, more easily

implemented diagnostic test weighed against potential
2357 in non-inferiority analysis group
missed diagnoses. We used Tango’s score method for 618 with a positive reference standard
matched pairs to derive CIs for differences in sensitivity 1739 with a negative reference standard
between comparator and index tests.20 Participants without a
determinate result on both the index and comparator test Figure 2: Participant selection. *One participant met more than one exclusion
were omitted from non-inferiority analyses. In an explora- criterion
MGIT=mycobacterial growth indicator tube liquid culture.
tory post-hoc analysis, Ultra semiquantitative trace results
were considered negative for M tuberculosis. There
was no prespecified non-inferiority margin for specificity were included in analyses. 242 (8%) participants were
comparisons. excluded because fewer than three sputum specimens were
Sample size calculations were done by Monte Carlo simu- collected and 64 (2%) were excluded because their reference
lation, conservatively assuming a moderate correlation of standard cultures were contaminated. Among 2600 partic-
0⋅5 between tests when testing samples from the same par- ipants analysed, 2357 (91%) had determinate results
ticipant using two different methods. We generated 10 000 for Ultra-Sp1-raw, Ultra-Sp2-pellet, and MGIT-Sp2-pellet
correlated binary data sets for each simulation, using a variety and were included in non-inferiority analyses. Among
of input test sensitivity parameters. The final sample size was 2357 participants in the non-inferiority analysis population,
selected to show superiority in 80% or more of the simulated 877 (37%) were HIV-positive and 525 (22%) reported a
datasets, with parameter estimates of 88% for Ultra sensi- history of previous treated tuberculosis (table 1). 618 (26%)
tivity and of 90% for MGIT sensitivity for case detection participants were positive for M tuberculosis by the reference
and non-inferiority margin of 5%. We calculated that 671 standard, including 477 (77%) of 618 who were positive
reference standard positive participants were required. using sputum smear microscopy.
In the primary analysis, the sensitivity of Ultra performed
Role of the funding source directly on sputum 1 (Ultra-Sp1-raw) was non-inferior to
The funder of the study had no role in study design, data that of sputum 2 MGIT culture (MGIT 91⋅1% vs Ultra
collection, data analysis, data interpretation, or writing of the 91⋅9%; difference –0⋅8 percentage points; 95% CI –2⋅8 to
report. 1⋅1; figure 3A). The sensitivity of Ultra performed on
sputum 2 pellet (Ultra-Sp2-pellet) was also non-inferior to
Results that of sputum 2 MGIT (MGIT 91⋅1% vs Ultra 91⋅9%;
Between Feb 18, 2016, and Dec 4, 2019, we screened difference –0⋅8 percentage points; 95% CI –2⋅7 to 1⋅0).
3489 people for potential enrolment and enrolled 2906 Sensitivity differences between Ultra and MGIT test pairs
(83%; figure 2). Among enrolled participants, 2600 (89%) were also assessed by baseline sputum smear microscopy

4 www.thelancet.com/microbe Vol ▪ ▪ 2024

 Articles

status, HIV status, and previous tuberculosis treatment

Eligible Determinate Non-inferiority
status. Sensitivity point estimates were slightly higher for population reference standard analysis population
Ultra tests than for MGIT in all subgroups except for (n=2906) population (n=2600) (n=2357)
sputum smear-negative participants (figure 3B). Median age (IQR), years 37 (19) 37 (19) 38 (20)
In post-hoc analyses in which Ultra semiquantitative Sex
results of trace were considered negative for M tuberculosis, Female 1213 (42%) 1094 (42%) 984 (42%)
the sensitivity of Ultra-Sp1-raw (89⋅2%) remained Male 1692 (58%) 1506 (58%) 1373 (58%)
non-inferior to that of MGIT culture (difference Not reported 1 (0%) 0 (0%) 0 (0%)
1⋅9 percentage points; 95% CI –0⋅1 to 4⋅1). The sensi- Enrolment country
tivity of Ultra-Sp2-pellet was not non-inferior to that of Belarus 52 (2%) 52 (2%) 50 (2%)
MGIT (difference 3⋅7 percentage points; 1⋅8 to 5⋅9; Brazil 174 (6%) 160 (6%) 152 (6%)
table 2). When the results were further stratified by China 37 (1%) 35 (1%) 35 (1%)
previous tuberculosis history (table 2), reclassification of Georgia 375 (13%) 369 (14%) 341 (14%)
trace Ultra results as negative resulted in a decrement India 247 (8%) 190 (7%) 176 (7%)
in Ultra-Sp2-pellet sensitivity from 92⋅1% (70 of 76; 83⋅8 Kenya 494 (17%) 416 (16%) 391 (17%)
to 96⋅3) to 77⋅6% (59 of 76; 67⋅1 to 85⋅5), whereas South Africa 901 (31%) 793 (31%) 716 (30%)
Ultra-Sp1-raw sensitivity was less affected. Uganda 626 (22%) 585 (23%) 496 (21%)
Compared with the sensitivity of sputum 2 MGIT, the HIV status
sensitivity of Xpert-Sp1-raw was not non-inferior (92⋅5% HIV-positive 1123 (39%) 989 (38%) 877 (37%)
[529 of 572; 95% CI 90⋅0 to 94⋅4] for MGIT vs 87⋅8% HIV-negative 1654 (57%) 1521 (59%) 1397 (59%)
[502 of 572; 84⋅8 to 90⋅2] for Xpert-Sp1-raw; difference Unknown HIV status 129 (4%) 90 (3%) 83 (4%)
4⋅7 percentage points; 2⋅7 to 7⋅1) and the sensitivity History of previous treated tuberculosis
of sputum 2 Löwenstein–Jensen culture was inferior Yes 629 (22%) 573 (22%) 525 (22%)
(91⋅6% [545 of 595; 89⋅1 to 93⋅6] for MGIT vs 79⋅3% [472 of No 2272 (78%) 2024 (78%) 1829 (78%)
595; 75⋅9 to 82⋅4] for Löwenstein–Jensen; difference Unknown 5 (0%) 3 (0%) 3 (0%)
12⋅3 percentage points; 9⋅9 to 15⋅2). MRS results
In the non-inferiority analysis group, MGIT specificity MRS-positive for Mycobacterium tuberculosis 639 (22%) 639 (25%) 618 (26%)
was 97⋅1% (1688 of 1739; 95% CI 96⋅2 to 97⋅8). Ultra-Sp1-raw Any smear-positive 487 (17%) 487 (19%) 477 (20%)
specificity was 94⋅4% (1642 of 1739; 93⋅2 to 95⋅4) with a No smear-positive 150 (5%) 150 (6%) 139 (6%)
difference of 2⋅7 percentage points (1⋅7 to 3⋅8) versus MGIT. MRS-negative for M tuberculosis 1961 (67%) 1961 (75%) 1739 (74%)
Ultra-Sp2-pellet specificity was 91⋅0% (1583 of 1739; 89⋅6 to Non-determinate due to contamination 306 (11%) NA NA
92⋅3), with a difference of 6⋅1 percentage points (4⋅9 to 7⋅5). In or third specimen not collected
post-hoc analyses in which Ultra trace results were consid- MRS=microbiological reference standard. NA=not applicable.
ered negative for M tuberculosis, Ultra-Sp1-raw specificity was
Table 1: Characteristics of the study population
95⋅7% (1664/1739; 94⋅6 to 96⋅6) with a difference of
1⋅4 percentage points (0⋅5 to 2⋅4) versus MGIT, and Ultra-
Sp2-pellet specificity was 96⋅2% (1672 of 1739; 95⋅1 to 97⋅0) Discussion
with a difference 1⋅0 percentage point (0⋅1 to 1⋅9) versus The sensitivity of one Ultra test was non-inferior to that of
MGIT (table 2). one MGIT liquid culture for detection of M tuberculosis in
The non-inferiority analysis required determinate expectorated sputum specimens. Non-inferiority was
results for both index and comparator tests in a pair and, demonstrated when MGIT and Ultra tests were performed
therefore, excluded some determinate results. Accord- on the same resuspended processed sputum pellet and
ingly, we calculated the sensitivity and specificity for each when Ultra tests were performed directly on a different raw,
test type considered independently (appendix p 2). The unprocessed sputum specimen. In the primary analysis,
sensitivity was 91⋅5% (581 of 635; 95% CI 89⋅1 to 93⋅4) there was no evidence that Ultra test sensitivity was less than
for Ultra-Sp1-raw, 91⋅5% (581 of 635; 89⋅1 to 93⋅4) for from that of MGIT.
Ultra-Sp2-pellet, and 90⋅9% (568 of 625; 88⋅4 to 92⋅9) for Results were consistent when stratified by participant
MGIT-Sp2-pellet. The specificity was 94⋅3% (1804 of sputum smear microscopy status and by HIV status. Ultra
1914; 93⋅1 to 95⋅2) for Ultra-Sp1-raw, 91⋅1% (1759 of sensitivity was lower in participants who were smear-
1931; 89⋅7 to 92⋅3) for Ultra-Sp2-pellet, and 97⋅2% (1758 negative versus smear-positive, supporting an influence of
of 1808; 96⋅4 to 97⋅9) for MGIT-Sp2-pellet. Among tests airway bacillary burden on test sensitivity, as has been pre-
initiated, a final result of non-determinate occurred in viously reported.6,7 Our results extend this observation to
0⋅2% (four of 2553) of Ultra-Sp1-raw test, 0⋅3% (eight of MGIT culture, for which sensitivity was substantially lower
2574) of Ultra-Sp2-pellet test, and 6⋅4% (167/2600) of among participants who were smear-negative compared
MGIT-Sp2-pellet tests. Contamination rates for MGIT with those who were smear-positive. MGIT and Ultra sen-
and Löwenstein–Jensen were between 6⋅1% and 9⋅9% sitivities were slightly lower in participants with HIV versus
(appendix p 3). those without HIV, a finding that might also reflect

www.thelancet.com/microbe Vol ▪ ▪ 2024 5

 Articles

A
Index test Index test sensitivity MGIT Sp2-pellet comparator Sensitivity difference: comparator Sensitivity difference:
(95% CI), %; n/n test sensitivity (95% CI), %; n/n minus index (95% CI), % comparator minus index (95% CI)

Ultra-Sp1-raw 91·9% (89·5 to 93·8); 568/618 91·1% (88·6 to 93·1); 563/618 –0·8% (–2·8 to 1·1)
Ultra-Sp2-pellet 91·9% (89·5 to 93·8); 568/618 91·1% (88·6 to 93·1); 563/618 –0·8% (–2·7 to 1·0)
Xpert-Sp1-raw 87·8% (84·8 to 90·2); 502/572 92·5% (90·0 to 94·4); 529/572 4·7% (2·7 to 7·1)
Löwenstein–Jensen Sp2-pellet 79·3% (75·9 to 82·4); 472/595 91·6% (89·1 to 93·6); 545/595 12·3% (9·9 to 15·2)

–4 –2 0 2 4 6 8 10 12 14 16
Higher sensitivity of MGIT

B
Index test Subgroup Index test sensitivity (95% MGIT Sp2-pellet comparator Sensitivity difference: Sensitivity difference: comparator minus index
CI), %; n/n test sensitivity (95% comparator minus with 95% CI
CI), %; n/n index (95% CI), %

Ultra Sp1-raw Smear-negative 66·2% (58·0 to 73·5); 92/139 66·9% (58·7 to 74·2); 93/139 0·7% (–6·6 to 8·1)
Smear-positive 99·4% (98·2 to 99·8); 474/477 98·1% (96·5 to 99·0); 468/477 –1·3% (–2·8 to –0·1)
HIV-negative 92·3% (89·2 to 94·5); 358/388 92·0% (88·9 to 94·3); 357/388 –0·3% (–2·5 to 1·9)
HIV-positive 91·7% (86·8 to 94·9); 166/181 89·0% (83·6 to 92·7); 161/181 –2·8% (–7·7 to 1·9)
Previous tuberculosis treatment 90·8% (82·2 to 95·5); 69/76 88·2% (79·0 to 93·6); 67/76 –2·6% (–9·9 to 3·7)
No previous tuberculosis treatment 92·0% (89·5 to 94·0); 497/540 91·5% (88·8 to 93·6); 494/540 –0·6% (–2·7 to 1·5)
Ultra Sp2-pellet Smear-negative 66·9% (58·7 to 74·2); 93/139 66·9% (58·7 to 74·2); 93/139 0·0% (–7·0 to 7·0)
Smear-positive 99·4% (98·2 to 99·8); 474/477 98·1% (96·5 to 99·0); 468/477 –1·3% (–2·7 to –0·5)
HIV-negative 93·0% (90·1 to 95·2); 361/388 92·0% (88·9 to 94·3); 357/388 –1·0% (–3·1 to 0·8)
HIV-positive 90·1% (84·8 to 93·6); 163/181 89·0% (83·6 to 92·7); 161/181 –1·1% (–5·9 to 3·5)
Previous tuberculosis treatment 92·1% (83·8 to 96·3); 70/76 88·2% (79·0 to 93·6); 67/76 –4·0% (–11·7 to 2·5)
No previous tuberculosis treatment 91·9% (89·2 to 93·9); 96/540 91·5% (88·8 to 93·6); 494/540 –0·4% (–2·3 to 1·5)

–12 –10 –8 –6 –4 –2 0 2 4 6 8 10 12
Higher sensitivity of MGIT

Figure 3: Results of pairwise comparisons of sensitivity between the comparator MGIT culture test and index tests for the non-inferiority analysis population
The dotted vertical line represents the margin of non-inferiority (5⋅0%). (A) Primary analysis. (B) Subgroup analyses. MGIT=mycobacterial growth indicator tube liquid culture.

differences in airway bacillary burden.21,22 The sensitivities important practical implications, especially in settings with
of MGIT-Sp2-pellet, Ultra-Sp2-pellet, and Ultra-Sp1-raw low resources for testing. Under a hypothetical testing
were mostly similar to each other across clinically import- scenario, in which only one sputum specimen was tested,
ant subgroups of participants living with HIV and partic- the use of Ultra would be expected to detect slightly more
ipants with low bacillary burden in sputum. These findings pulmonary tuberculosis cases than would the use of MGIT
provide reassurance that Ultra-based diagnostic testing culture given Ultra’s comparable sensitivity and higher
algorithms maintain sensitivity in settings where HIV is proportion of determinate results. Non-determinate
prevalent. sputum 2 MGIT results were likely due to contamination
We selected a non-inferiority study design due to the high rates of 6⋅6%. A contamination rate of 5–8% is considered
intrinsic sensitivity of the comparator MGIT culture, such the benchmark rate that balances the competing risks of
that a superiority trial would not be practicable, and because false-negative cultures due to loss of mycobacterial viability
Ultra has attributes that might be sufficiently favourable to during sputum decontamination procedures with over-
outweigh a potential small reduction in sensitivity. To our growth of upper airway flora due to insufficient sputum
knowledge this is the first study to use a non-inferiority decontamination.17
design in tuberculosis diagnostics, but might become Specificity point estimates for Ultra and Xpert were lower
increasingly necessary given the emergence of other point- in our study than in some other recent studies.6,7
of-care nucleic acid amplification tests for tuberculosis A contributing factor to the lower specificity point esti-
diagnosis. We selected the non-inferiority margin based on mates in our study was the analysis approach, in which
informal input from tuberculosis clinicians, although a results of only one sputum specimen (ie, sputum 3) were
Delphi-type approach would have been more rigorous. considered for reference standard classification of each
Nonetheless, our conclusion of non-inferiority of sensitivity participant. We used this approach to designate the sputum
is not highly influenced by the absolute margin. 2 MGIT test as the comparator test, given that our overall
The proportion of participants with a determinate test study hypothesis focused on test sensitivity and to maintain
result was higher for Ultra than for MGIT, which has an overall testing approach in which no more than three

6 www.thelancet.com/microbe Vol ▪ ▪ 2024

www.thelancet.com/microbe Vol ▪ ▪ 2024

Non-inferiority analysis group No previous history of tuberculosis Previously treated tuberculosis

Index test MGIT-Sp2-pellet Sensitivity difference: Index test MGIT-Sp2-pellet Sensitivity difference: Index test MGIT-Sp2-pellet Sensitivity difference:
sensitivity comparator test comparator minus sensitivity comparator test comparator minus sensitivity comparator test comparator minus
sensitivity index sensitivity index sensitivity index
Sensitivity
Ultra-Sp1-raw, trace as 91⋅9% 91⋅1% (88⋅6 to 93⋅1); –0⋅8% (–2⋅8 to 1⋅1) 92⋅0% 91⋅5% (88⋅8 to 93⋅6); –0⋅6% (–2⋅7 to 1⋅5) 90⋅8% 88⋅2% (79⋅0 to 93⋅6); –2⋅6% (–9⋅9 to 3⋅7)
Mycobacterium (89⋅5 to 93⋅8); 563/618 (89⋅5 to 94⋅0); 494/540 (82⋅2 to 95⋅5); 67/76
tuberculosis-positive 568/618 497/540 69/76
Ultra-Sp1-raw, trace as 89⋅2% .. 1⋅9% (–0⋅1 to 4⋅1) 89⋅3% .. 2⋅2% (0⋅1 to 4⋅6) 88⋅2% .. 0⋅0% (–7⋅7 to 7⋅7)
M tuberculosis-negative (86⋅5 to 91⋅4); (86⋅4 to 91⋅6); (79⋅0 to 93⋅6);
551/618 482/540 67/76
Ultra-Sp2-pellet, trace as 91⋅9% .. –0⋅8% (–2⋅7 to 1⋅0) 91⋅9% .. –0⋅4% (–2⋅3 to 1⋅5) 92⋅1% .. –4⋅0% (–11⋅7 to 2⋅5)
M tuberculosis-positive (89⋅5 to 93⋅8); (89⋅2 to 93⋅9); (83⋅8 to 96⋅3);
568/618 496/540 70/76
Ultra-Sp2-pellet, trace as 87⋅4% .. 3⋅7% (1⋅8 to 5⋅9) 88⋅7% .. 2⋅8% (0⋅8 to 5⋅0) 77⋅6% .. 10⋅5% (3⋅0 to 20⋅0)
M tuberculosis-negative (84⋅5 to 90⋅0); (85⋅8 to 91⋅1); (67⋅1 to 85⋅5);
540/618 479/540 59/76
Specificity
Ultra-Sp1-raw, trace as 94⋅4% 97⋅1% (96⋅2 to 97⋅8); 2⋅7% (1⋅7 to 3⋅8) 95⋅0% 96⋅7% (95⋅6 to 97⋅6); 1⋅7% (0⋅6 to 2⋅9) 92⋅7% 98⋅2% (96⋅5 to 99⋅1); 5⋅6% (3⋅2 to 8⋅4)
M tuberculosis-positive (93⋅2 to 95⋅4); 1688/1739 (93⋅7 to 96⋅1); 1247/1289 (89⋅9 to 94⋅7); 441/449
1642/1739 1225/1289 416/449
Ultra-Sp1-raw, trace as 95⋅7% .. 1⋅4% (0⋅5 to 2⋅4) 96⋅8% .. –0⋅1% (–1⋅1 to 0⋅9) 95⋅1% .. 3⋅1% (1⋅1 to 5⋅5)
M tuberculosis-negative (94⋅6 to 96⋅6); (95⋅7 to 97⋅7); (92⋅7 to 96⋅7);
1664/1739 1248/1289 427/449
Ultra-Sp2-pellet, trace as 91⋅0% .. 6⋅1% (4⋅9 to 7⋅5) 91⋅3% .. 5⋅4% (4⋅1 to 7⋅0) 90⋅2% .. 8⋅0% (5⋅4 to 11⋅1)
M tuberculosis-positive (89⋅6 to 92⋅3); (89⋅7 to 92⋅7); (87⋅1 to 92⋅6);
1583/1739 1177/1289 405/449
Ultra-Sp2-pellet, trace as 96⋅2% .. 1⋅0% (0⋅1 to 1⋅9) 96⋅4% .. 0⋅3% (–0⋅7 to 1⋅3) 95⋅3% .. 2⋅9% (0⋅8 to 5⋅3)
M tuberculosis-negative (95⋅1 to 97⋅0); (95⋅3 to 97⋅3); (93⋅0 to 96⋅9);
1672/1739 1243/1289 428/449
Data are in % (95% CI); n/N or % (95% CI). MGIT=mycobacterial growth indicator tube liquid culture.

Table 2: Sensitivity and specificity of Ultra vs MGIT, by interpretation of trace Ultra results and history of prior treated tuberculosis

Articles
7
 Articles

sputum specimens were required from each participant. because fewer than three sputum specimens were
A reference standard that includes results of two or more obtained. This analysis approach could have led to
sputum specimens reduces the effect of sputum-to-sputum selection bias if people with paucibacillary disease were
bacillary burden heterogeneity on participant classification less likely to produce three sputum specimens. The
since isolation of M tuberculosis from any of the corre- Ultra package insert recommends using a 3:1 ratio of
sponding cultures would classify the participant as sample reagent to specimen (volume) when testing
reference-standard positive. This reference standard also resuspended processed sputum pellets. This is a con-
reduces the number of participants excluded from the venience recommendation to help ensure an adequate
analysis due to a non-determinate reference standard. specimen volume for Ultra testing in the event that the
In the current study, the sputum 2 MGIT specificity of residual pellet volume is low. We used a 2:1 ratio of
96⋅6% provides some insight into the effect of specimen-to- sample reagent to specimen pellet volume based on
specimen heterogeneity, since 51 (2⋅9%) of 1739 specimens previous work showing superior sensitivity versus a 3:1
that were negative for M tuberculosis by the culture-based dilution.32 Our study was not designed to identify the
sputum 3 reference standard were positive for optimal number of sputum specimens per person that
M tuberculosis by sputum 2 MGIT culture. should be tested using Ultra to optimise the diagnostic
Our results extend previous observations that Ultra yield of a clinical testing algorithm. We acknowledge
detects M tuberculosis in some specimens from which that tuberculosis diagnostic guidelines recommend test-
M tuberculosis cannot be cultivated using routine ing more than one sputum sample and that Ultra test-
methods.6,23–28 The biological underpinnings for such dis- ing on processed sputum pellets typically would not be
cordant results have not been identified; however, work in done at point-of-care. With respect to study design,
this area could improve understanding of the pathobiology we used the MGIT culture method as a comparator
of tuberculosis when considered as a disease spectrum. (specimen 2) and a reference standard (specimen 3) and
When tuberculosis is viewed as a dichotomous condition, we mitigated potential bias by not including specimen 2
possible explanations for discordant results between Ultra MGIT results in the reference standard.
and culture include false-positive Ultra results due to dead Our findings provide new, clinically relevant informa-
bacilli or contamination and false-negative cultures.26 MGIT tion about the head-to-head comparative diagnostic
culture is an imperfect reference standard that can give sensitivity of Ultra and MGIT culture for the detection
false-negative results due to the killing of M tuberculosis of M tuberculosis in sputum specimens. Results should
during the sputum decontamination process or the help policy makers and tuberculosis programmes calcu-
presence of differentially cultivatable bacilli that do not late the relative merits and drawbacks of diagnostic test-
grow in routine culture.14,15 These scenarios would most ing strategies that incorporate these assays. Highly
probably occur with a low, rather than high, sputum sensitive, rapid molecular approaches for tuberculosis
bacillary burden. Consistent with this, Ultra-positive and case detection combined with advances in genotypic
culture-negative rates as high as 48–65% have been approaches to drug resistance detection have the potential
observed in high prevalence settings for tuberculosis to replace culture.
where people were tested regardless of symptoms,
whereas concordance between Ultra and culture has been Contributors
YLX, SGS, MLJ, JJE, DA, and SED designed the study. YLX, CE, EN, NH,
substantially higher in people with pulmonary tubercu- WZ, RO, IA, LN, JvH, KDM, KA, SK, JEK, CR, AS, WS, MH, RD, XL, and
losis symptoms, as in our study.26,27,29,30 As expected, the SED oversaw the trial conduct. YLX, CE, KA, JEK, and SED analysed data
designation of Ultra trace results as negative increased and developed the first manuscript draft. All authors contributed to data
specificity but reduced sensitivity, although Ultra-Sp2-raw collection, interpretation of data, and revision of the manuscript. All
authors had full access to all the data in the study and had final
sensitivity remained non-inferior to that of the MGIT
responsibility for the decision to submit for publication.
comparator. From a tuberculosis programme perspective,
this trade-off between sensitivity and specificity is weighted Declaration of interests
against factors, including the prevalence of tuberculosis, the SGS was employed by FIND during conduct of the study. FIND is a not-
prevalence of comorbidities that affect tuberculosis mor- for-profit foundation that supports the evaluation of publicly prioritised
tuberculosis assays and the implementation of WHO-approved (guidance
bidity and mortality, and the resources for close follow-up of
and prequalification) assays using donor grants. FIND has product
people with Ultra trace results who are on tuberculosis evaluation agreements with several private sector companies that design
treatment.31 diagnostics and related products for treatment of tuberculosis and other
A key study strength is the ability to compare the diseases. These agreements strictly define FIND’s independence and
sensitivities of Ultra and MGIT performed on aliquots neutrality with regard to the companies whose products get evaluated and
describe roles and responsibilities. DA receives income from licence
of the same specimen (ie, resuspended pellet from payments from Cepheid to Rutgers University and reports receiving
sputum 2). The diversity of study settings, inclusion of research contracts and research support from Cepheid. All other authors
people living with and without HIV, and heterogeneity declare no competing interests.
in sputum bacillary burden contribute to the general- Data sharing
isability of study findings. A study limitation is that 8% De-identified participant data can be accessed on reasonable request to the
of enrolled participants were excluded from analyses corresponding author.

8 www.thelancet.com/microbe Vol ▪ ▪ 2024

 Articles

Acknowledgments 15 Chengalroyen MD, Beukes GM, Gordhan BG, et al. Detection and
We thank the study participants, without whom this work would not quantification of differentially culturable tubercle bacteria in sputum
have been possible, and the clinical and laboratory teams, including from patients with tuberculosis. Am J Respir Crit Care Med 2016;
194: 1532–40.
CM Centner, at the participating trial sites. This study was funded by
16 Global Laboratory Initiative of the StopTB Partnership. Laboratory
the US National Institutes of Health (1R01AI129411, 1U01AI152084,
diagnosis of tuberculosis by sputum microscopy. Adelaide, SA:
K24AI104830; contract AI2008026). Xpert MTB/RIF and Xpert MTB/RIF SA Pathology, 2013.
Ultra cartridges were donated by Cepheid. Cepheid personnel had no role
17 Association of Public Health Laboratories. Essentials for the
in study design, implementation, analysis, manuscript writing, or decision mycobacteriology laboratory. https://www.aphl.org/programs/
to submit the findings for publication. The findings and conclusions in this infectious_disease/tuberculosis/TBCore/Mycobacterial_Culture-
report are those of the authors and do not necessarily represent the official WithNotes.pdf (accessed Feb 8, 2023).
position of the US National Institutes of Health. 18 Global Laboratory Initiative. Mycobacteriology laboratory manual.
Geneva: Global Laboratory Initiative, 2014.
References 19 Siddiqi SH, Rusch-Gerdes S. MGIT procedure manual for BACTEC
1 Boehme CC, Nabeta P, Hillemann D, et al. Rapid molecular MGIT 960 TB System. https://www.finddx.org/wp-content/
detection of tuberculosis and rifampin resistance. N Engl J Med 2010; uploads/2023/02/20061101_rep_mgit_manual_FV_EN.pdf
363: 1005–15. (accessed March 6, 2023).
2 Horne DJ, Kohli M, Zifodya JS, et al. Xpert MTB/RIF and Xpert 20 Tango T. Equivalence test and confidence interval for the difference
MTB/RIF Ultra for pulmonary tuberculosis and rifampicin in proportions for the paired-sample design. Stat Med 1998;
resistance in adults. Cochrane Database Syst Rev 2019; 6: CD009593. 17: 891–908.
3 GeneXpert. Xpert MTB/RIF. July, 2020. https://www.cepheid.com/ 21 Elliott AM, Halwiindi B, Hayes RJ, et al. The impact of human
Package%20Insert%20Files/Xpert-MTB-RIF-ENGLISH-Pakage- immunodeficiency virus on presentation and diagnosis of tuberculosis
Insert-301-1404-Rev-G.pdf (accessed June 28, 2022). in a cohort study in Zambia. J Trop Med Hyg 1993; 96: 1–11.
4 WHO. Xpert MTB/RIF assay for the diagnosis of pulmonary and 22 Elliott AM, Namaambo K, Allen BW, et al. Negative sputum smear
extrapulmonary TB in adults and children: policy update. Geneva: results in HIV-positive patients with pulmonary tuberculosis in
World Health Organization, 2014. Lusaka, Zambia. Tuber Lung Dis 1993; 74: 191–94.
5 Chakravorty S, Simmons AM, Rowneki M, et al. The new Xpert 23 Esmail A, Tomasicchio M, Meldau R, Makambwa E, Dheda K.
MTB/RIF Ultra: improving detection of Mycobacterium tuberculosis Comparison of Xpert MTB/RIF (G4) and Xpert Ultra, including trace
and resistance to rifampin in an assay suitable for point-of-care readouts, for the diagnosis of pulmonary tuberculosis in a TB and
testing. MBio 2017; 8: e00812–17. HIV endemic setting. Int J Infect Dis 2020; 95: 246–52.
6 Dorman SE, Schumacher SG, Alland D, et al. Xpert MTB/RIF Ultra 24 Amedeo A, Beci G, Giglia M, et al. Evaluation of trace calls by Xpert
for detection of Mycobacterium tuberculosis and rifampicin resistance: MTB/RIF ultra for clinical management in low TB burden settings.
a prospective multicentre diagnostic accuracy study. Lancet Infect Dis PLoS One 2022; 17: e0272997.
2018; 18: 76–84. 25 Mishra H, Reeve BWP, Palmer Z, et al. Xpert MTB/RIF Ultra and
7 Zifodya JS, Kreniske JS, Schiller I, et al. Xpert Ultra versus Xpert Xpert MTB/RIF for diagnosis of tuberculosis in an HIV-endemic
MTB/RIF for pulmonary tuberculosis and rifampicin resistance in setting with a high burden of previous tuberculosis: a two-cohort
adults with presumptive pulmonary tuberculosis. diagnostic accuracy study. Lancet Respir Med 2020; 8: 368–82.
Cochrane Database Syst Rev 2021; 2: CD009593. 26 Berhanu RH, Lebina L, Nonyane BAS, et al. Yield of facility-based
8 Kay AW, González Fernández L, Takwoingi Y, et al. Xpert MTB/RIF targeted universal testing for tuberculosis with Xpert and
and Xpert MTB/RIF Ultra assays for active tuberculosis and mycobacterial culture in high-risk groups attending primary care
rifampicin resistance in children. Cochrane Database Syst Rev 2020; facilities in South Africa. Clin Infect Dis 2023; 76: 1594–603.
8: CD013359. 27 Moyo S, Ismail F, Van der Walt M, et al. Prevalence of
9 Kohli M, Schiller I, Dendukuri N, et al. Xpert MTB/RIF Ultra and bacteriologically confirmed pulmonary tuberculosis in South Africa,
Xpert MTB/RIF assays for extrapulmonary tuberculosis and 2017–19: a multistage, cluster-based, cross-sectional survey.
rifampicin resistance in adults. Cochrane Database Syst Rev 2021; Lancet Infect Dis 2022; 22: 1172–80.
1: CD012768. 28 Mekkaoui L, Hallin M, Mouchet F, et al. Performance of Xpert MTB/
10 WHO. Molecular assays intended as initial tests for the diagnosis of RIF Ultra for diagnosis of pulmonary and extra-pulmonary
pulmonary and extrapulmonary TB and rifampicin resistance in tuberculosis, one year of use in a multi-centric hospital laboratory in
adults and children: rapid communication. Geneva: World Health Brussels, Belgium. PLoS One 2021; 16: e0249734.
Organization, 2020. 29 Kendall EA, Kitonsa PJ, Nalutaaya A, et al. The spectrum of
11 Peres RL, Maciel ELN, Morais CG, et al. Comparison of two tuberculosis disease in an urban Ugandan community and its health
concentrations of NALC-NaOH for decontamination of sputum for facilities. Clin Infect Dis 2021; 72: e1035–43.
mycobacterial culture. Int J Tuberc Lung Dis 2009; 13: 1572–75. 30 Scott L, Da Silva P, Fyve K. Characterisation of South Africa’s Xpert
12 Pheiffer C, Carroll NM, Beyers N, et al. Time to detection of MTB/RIF Ultra “trace” laboratory results. Top Antivir Med 2020;
Mycobacterium tuberculosis in BACTEC systems as a viable alternative 28: 483.
to colony counting. Int J Tuberc Lung Dis 2008; 12: 792–98. 31 Kendall EA, Schumacher SG, Denkinger CM, Dowdy DW. Estimated
13 Mtafya B, Sabiiti W, Sabi I, et al. Molecular bacterial load assay clinical impact of the Xpert MTB/RIF Ultra cartridge for diagnosis of
concurs with culture on NaOH-induced loss of Mycobacterium pulmonary tuberculosis: a modeling study. PLoS Med 2017;
tuberculosis viability. J Clin Microbiol 2019; 57: e01992–18. 14: e1002472.
14 Mukamolova GV, Turapov O, Malkin J, Woltmann G, Barer MR. 32 Dharan NJ, Amisano D, Mboowa G, et al. Improving the sensitivity
Resuscitation-promoting factors reveal an occult population of of the Xpert MTB/RIF assay on sputum pellets by decreasing the
tubercle bacilli in sputum. Am J Respir Crit Care Med 2010; amount of added sample reagent: a laboratory and clinical evaluation.
181: 174–80. J Clin Microbiol 2015; 53: 1258–63.

www.thelancet.com/microbe Vol ▪ ▪ 2024 9