Shete et al.

BMC Infectious Diseases (2019) 19:268

https://doi.org/10.1186/s12879-019-3881-y

RESEARCH ARTICLE Open Access

Diagnostic accuracy of TB-LAMP for

pulmonary tuberculosis: a systematic
review and meta-analysis
Priya B. Shete1,2*† , Katherine Farr1,2†, Luke Strnad3, Christen M. Gray4 and Adithya Cattamanchi1,2

Abstract
Background: The need for a rapid, molecular test to diagnose tuberculosis (TB) has prompted exploration of TB-LAMP
(Eiken; Tokyo, Japan) for use in resource-limited settings. We conducted a systematic review to assess the accuracy of
TB-LAMP as a diagnostic test for pulmonary TB.
Methods: We analyzed individual-level data for eligible patients from all studies of TB-LAMP conducted between Jan
2012 and October 2015 to compare the diagnostic accuracy of TB-LAMP with that of smear microscopy and Xpert
MTB/RIF® using 3 reference standards of varying stringency. Pooled sensitivity and specificity and pooled differences in
sensitivity and specificity were estimated using random effects meta-analysis. Study quality was evaluated using
QUADAS-2.
Results: Four thousand seven hundred sixty individuals across 13 studies met eligibility criteria. Methodological quality
was judged to be low for all studies. TB-LAMP had higher sensitivity than sputum smear microscopy (pooled sensitivity
difference + 13·2, 95% CI 4·5–21·9%) and similar sensitivity to Xpert MTB/RIF (pooled sensitivity difference − 2·5, 95% CI
-8·0 to + 2·9) using the most stringent reference standard available. Specificity of TB-LAMP was similar to that of
sputum smear microscopy (pooled specificity difference − 1·8, 95% CI -3·8 to + 0·2) and Xpert MTB/RIF (pooled
specificity difference 0·5, 95% CI -0·9 to + 1·8).
Conclusions: From the perspective of diagnostic accuracy, TB-LAMP may be considered as an alternative test for
sputum smear microscopy. Additional factors such as cost, feasibility, and acceptability in settings that continue
to rely on sputum smear microscopy should be considered when deciding to adopt this technology. Xpert MTB/
RIF should continue to be preferred in settings where resource and infrastructure requirements are adequate and
where HIV co-infection or drug-resistance is of concern.
Keywords: Tuberculosis, Diagnostic testing, Molecular assay, Point of care

Background since been considerable investment in its scale-up [1].

Better diagnostics are essential for achieving global However, relatively high device and consumable costs,
tuberculosis (TB) elimination targets. In 2013, Xpert infrastructure requirements, and need for continuous
MTB/RIF® (Xpert) (Cepheid, Sunnyvale, CA, USA) be- instrument maintenance remain obstacles to use of
came the first molecular TB test endorsed by the Xpert as a point-of-care test in peripheral health cen-
World Health Organization (WHO), and there has ters where the majority of TB patients initially
present for care.
To expand the availability of molecular testing for TB,
* Correspondence: Priya.shete@ucsf.edu
†
Priya B. Shete and Katherine Farr contributed equally to this work.
Eiken Chemical Co., Ltd. developed a commercial ver-
1
Division of Pulmonary and Critical Care Medicine, University of California sion of loop-mediated isothermal amplification (LAMP),
San Francisco and Zuckerberg San Francisco General Hospital, 1001 Potrero a technique in which nucleic acid amplification occurs
Avenue, 5K1, San Francisco, CA 94110, USA
2
Curry International Tuberculosis Center, University of California San
under isothermal conditions. LAMP is based on
Francisco, San Francisco, CA, USA auto-cycling, strand displacement DNA synthesis
Full list of author information is available at the end of the article

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 2 of 11

performed by a DNA polymerase with high strand Study and participant selection
displacement activity and two specially designed We included all studies which 1) evaluated the Eiken
inner and two outer primers [2, 3]. Eiken’s Loopamp TB-LAMP kit on sputum samples from adult presump-
MTBC Detection Kit (TB-LAMP) targets the gyrB tive TB patients; 2) were conducted in an intermediate
and IS regions of the Mycobacterium tuberculosis or high TB burden country as defined by WHO; and 3)
(MTB) complex genome. Detection of amplified were conducted after January 1, 2012 using the final
product is based on turbidity visualized with the protocol, training, and TB-LAMP kits approved by
naked eye or under ultraviolet (UV) light after 15– Eiken. We excluded studies that 1) did not exclude pa-
60 min [4–7]. tients on TB treatment within 60 days of enrollment; 2)
To inform WHO guideline development, we con- did not perform speciation testing to confirm presence
ducted a systematic review of studies evaluating the of Mycobacterium tuberculosis (MTB) complex in posi-
diagnostic accuracy of TB-LAMP using the latest tive cultures; or 3) performed TB-LAMP on frozen
assay kit and protocol. The primary objective was to specimens.
evaluate the diagnostic accuracy of TB-LAMP if used Authors of eligible studies provided individual partici-
as an alternative test for sputum smear microscopy pant data. We excluded individual participants who were
among adults suspected of having pulmonary TB. In 1) less than 18 years of age; 2) did not have results of
addition, we sought to determine the diagnostic ac- speciation testing for MTB; 3) had a positive culture but
curacy of TB-LAMP if used as an alternative test for speciation testing identified only non-tuberculous myco-
microscopy among adults with HIV infection or as an bacteria (NTM); 4) had a documented history of prior
add-on test for adults with negative sputum smear TB; 5) had TB-LAMP testing performed on non-sputum
microscopy results, and the proportion of indeter- samples; 6) had TB-LAMP testing done with a total
minate/invalid TB-LAMP results. reaction volume of < 25 μL; or 7) could not be classified
as TB-positive or TB-negative based on the reference
standard definitions described below. When comparing
Methods
TB-LAMP to Xpert, we also excluded individual
We followed standard guidelines and methods for sys-
participants for whom Xpert was performed on frozen
tematic reviews and meta-analyses of diagnostic tests
samples or valid results were unavailable for both
[8]. We developed four PICO style research questions to
TB-LAMP and Xpert.
inform this review. First, what is the diagnostic accuracy
of TB-LAMP for detection of pulmonary TB in
Quality assessment
reference to mycobacterial culture if used as an alterna-
We used the Assessment of Diagnostic Accuracy Studies
tive test for sputum smear microscopy among all adults
(QUADAS-2) tool to assess the methodological quality
and among HIV-infected adults? Second, what is the
of eligible studies [9]. Specific yes/no signaling questions
diagnostic accuracy of TB-LAMP for detection of pul-
were tailored for each QUADAS-2 domain.
monary TB in reference to mycobacterial culture if used
as an add-on test among sputum smear-negative adults?
Index tests
Third, what is the difference in diagnostic accuracy be-
Studies recorded LAMP and Xpert results as negative,
tween TB-LAMP and Xpert for detection of pulmonary
positive or indeterminate/invalid in accordance with
TB in reference to mycobacterial culture among all
manufacturer recommendations. We standardized spu-
adults? And finally, what is the proportion of indeter-
tum smear microscopy results across studies by consid-
minate/invalid results when TB-LAMP is used to detect
ering only direct ZN (Ziehl-Neelsen) and direct FM
pulmonary TB among all adults and among
(fluorescence microscopy) results, considering only the
HIV-infected adults?
first two smear results if more than two direct ZN or
direct FM results were available, and defining patients to
Search strategy be sputum smear-positive if ≥1 acid-fast bacillus (AFB)
To perform a study of TB-LAMP, investigators must was identified in any sputum smear.
order kits directly from Eiken. Therefore, a list of such
studies was requested from Eiken. To confirm the list Reference standard
provided was complete, we performed a search in We used three hierarchical mycobacterial culture-based
Google Scholar and PubMed using the terms “TB reference standards to account for differences in the
LAMP”, “TB-LAMP”, and “tuberculosis LAMP”. To number of cultures performed and results available. Pa-
meet the deadline for WHO guideline development in tients were classified as having TB if one or more cul-
January 2016, we only included studies completed by tures were positive and confirmed to be MTB complex
October 1, 2015. by speciation testing. Patients were classified as not
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 3 of 11

having TB if there were no positive cultures and at least 1) estimates along with 95% confidence regions in ROC
two negative cultures performed on different sputum sam- space. Pooled differences in sensitivity/specificity between
ples (Standard 1); 2) two negative cultures performed on TB-LAMP and Xpert and pooled proportion of indeter-
one sputum sample (Standard 2); or 3) one negative cul- minate/invalid TB-LAMP results were generated using
ture (Standard 3). The three reference standards random effects meta-analysis. For the primary review
allowed for a trade-off between yield of TB diagnosis question of TB-LAMP accuracy if used as an alernative
(highest with Standard 1 and lowest with Standard 3) test for smear microscopy, we explored potential reasons
and number of studies/participants included in each for heterogeneity by performing sub-group analyses based
analysis (lowest with Standard 1 and highest with on the health system level (reference lab, microscopy cen-
Standard 3). ter, or hospital-affiliated clinics) at which the study was
conducted and study quality (high-quality studies across
Statistical analysis all domains and within each domain of QUADAS-2).
For all review questions, we assessed heterogeneity
visually with forest plots and statistically with the χ2 Results
test for heterogeneity and the I2 test of inconsistency Of 20 studies identified by Eiken at the time of this
[10, 11]. When four or more studies were available, review, 13 met criteria for inclusion (Fig. 1): four
we generated pooled summary estimates of sensitivity/ evaluation (EVAL) studies were conducted by the
specificity using hierarchical summary receiver operating Foundation for Innovative New Diagnostics (FIND) in
characteristic (HSROC) analysis. We plotted these reference labs; one demonstration (DEMO) study was

Fig. 1 Study and participant selection flow diagram. Of 20 potentially eligible studies, 13 met study-level eligibility criteria. The 13 eligible studies
included 5099 participants, of whom 339 (7%) did not meet participant-level eligibility criteria. Of the 4760 eligible participants, 1810 (38%) were
included in the analysis for reference standard 1, 3110 (65%) for reference standard 2, and 4596 (97%) for reference standard 3. Abbreviations: TB,
Tuberculosis; MTB, Mycobacterium tuberculosis; μL, microliter
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 4 of 11

conducted by FIND in a peripheral microscopy cen- with Standards 2 and 3 (Additional file 1: Figure S1).
ter; seven studies were commissioned by FIND Visual inspection of forest plots indicated heterogeneity
through a request for applications (RFA); and one in specificity estimates was less than for sensitivity esti-
study was sponsored by Eiken [12–14]. Authors of in- mates, but was still significant (I2 61–78%, p < 0·03 for
cluded studies submitted individual data for 5099 par- all reference standards). Pooled specificity of TB-LAMP
ticipants, of whom 4760 were eligible for analysis ranged from 97·7% (95% CI 96·1–98·7) when using
(Additional file 1: Table S1). Standard 3 to 98·1% (95% CI 95·7–99·2) when using
Of the included studies, four were conducted at Standard 1 (Table 2 and Additional file 1: Figure S2).
reference laboratories, six at hospital and/or TB-LAMP had similar specificity to sputum smear mi-
university-affiliated clinics, and three at peripheral mi- croscopy, with specificity difference ranging from − 1·8%
croscopy centers (Table 1). Study participants were (95% CI -3·8–0·2) when using Standard 1 to − 1·3% (95%
majority male with a median age of 40 (IQR 29–54) CI -3·1% to + 0·4) when using Standard 3.
years. More than 10% of participants were known to Sub-group analyses did not reduce heterogeneity in
be HIV-positive in four of 13 studies (S. Africa EVAL, pooled sensitivity and specificity estimates. Pooled sensi-
Malawi RFA, Uganda RFA, Ivory Coast RFA). The pro- tivity was 78·0% (95% CI 68·9–85·0) among studies con-
portion of patients with culture-positive TB was 20– ducted in reference labs and 86·8% (95% CI 67·9–95·4)
40% in most studies, but was notably lower (8–15%) in among studies conducted in hospital-affiliated clinics,
three studies (India DEMO, India RFA, Vietnam RFA) and there was significant heterogeneity within both
and higher (66%) in one study (Vietnam EVAL). The sub-groups (I2 74—96%, p < 0·01 for all reference stan-
proportion of patients with smear-negative TB ranged dards) (Additional file 1: Table S3). Among studies rated
widely from 13 to 59%. as high-quality for patient selection, pooled sensitivity
was 84·2% (95% CI 71·1–92·0) and pooled specificity
Methodological quality 98·1% (95% CI 94·3–99·4), but there was significant het-
We considered overall risk of bias to be high due to erogeneity across studies (I2 83%, p < 0·001 for sensitivity
problems with the culture-based reference standard (all and I2 87%, p < 0·001 for specificity). Pooled estimates
13 studies), unclear patient selection (five studies), and could not be obtained for studies conducted at micros-
flow and timing concerns (eight studies) (Fig. 2). We copy centers and studies rated as high-quality for
found applicability concerns to be low for index tests reference standard because there were less than four
and the reference standard; however, we judged five studies in these sub-groups.
studies to have high applicability concerns for patient se-
lection because they were conducted at referral labora- TB-LAMP accuracy if used as an alternative test for smear
tories/centers or because enrollment involved screening microscopy in HIV-infected adults
of patients by a pulmonary specialist (Additional file 1: Pooled sensitivity of TB-LAMP was lower among
Table S2). HIV-infected adults than all adults, ranging from
63·8% (95% CI 49·0–76·4) with Standard 2 to 73·4
TB-LAMP accuracy if used as an alternative test for smear (95% CI 51·9–87·6) with Standard 3 (Table 2 and
microscopy Additional file 1: Figure S3). Pooled specificity was
Sensitivity of TB-LAMP in individual studies ranged low with Standard 3 (95·0, 95% CI 64·0–99·5) but
from 66 to 91% with Standard 1 (Fig. 3), 62–91% high with Standard 2 (98·8, 95% CI 85·1–99·9). There
with Standard 2 and 48–100% with Standard 3 was considerable heterogeneity in sensitivity (I2 86%,
(Additional file 1: Figure S1). We found significant p < 0·001) and specificity (I2 86%, p < 0·001) estimates
heterogeneity in sensitivity estimates, both from visual with Standard 3, but not Standard 2 (I2 54%, p = 0·09 for
inspection of forest plots and statistical testing (I2 sensitivity and I2 0%, p = 0·42 for specificity) (Fig. 3). There
72–94%, p < 0·003 for all reference standards). Pooled were insufficient studies (N = 2) to perform meta-analysis
sensitivity of TB-LAMP ranged from 77·7% (95% CI using Standard 1.
71·2–83·0) when using Standard 1 to 80·3% (95% CI
70·3–87·5) when using Standard 3 (Table 2 and TB-LAMP accuracy if used as an add-on test in smear-
Additional file 1: Figure S2). TB-LAMP was more negative adults
sensitive than sputum smear microcopy, with the sen- As expected, pooled sensitivity of TB-LAMP was lower
sitivity difference ranging from 7·1% (95% CI 1·4– among smear-negative adults than among all adults,
12·9) when using Standard 1 to 13·2% (95% CI 4·5– ranging from 40·3% (95% CI 27·9–54·0) with Standard 3
21·9) when using Standard 3. to 42·2% (95% CI 27·9–57·9) with Standard 2 (Table 2
Specificity of TB-LAMP in individual studies ranged and Additional file 1: Figure S4). Pooled specificity of
from 90 to 99% with Standard 1 (Fig. 3), and 90–100% TB-LAMP among smear-negative adults was similar to
Table 1 Study characteristics
Study Health system level Microscopy TB culture MGIT Tests done on stored Xpert specimen Median Age Female HIVa Culture- Smear-
type type contamination rate sputum type (IQR) positive TBb negative TBc
Evaluation Studies
Brazil Reference Lab Direct ZN 2x MGIT, 2x 0% Xpert Frozen 48 (35–60) 40% 0.4% 32% 25%
×2 LJ Culture Processed
Peru Reference Lab Direct ZN 2x MGIT, 2x 1·3% Xpert Fresh Processed 43 (28–56) 50% 1.0% 22% 42%
×2 LJ Culture
Shete et al. BMC Infectious Diseases

South Africa Reference Lab Direct ZN 2x MGIT, 2x 3·1% Xpert Fresh Processed 39 (29–47) 34% 35% 26% 51%
×2 LJ Culture
Vietnam Reference Lab Direct ZN 2x MGIT, 2x 0% Xpert Frozen 39 (26–50) 30% 2% 66% 42%
×2 LJ Culture Processed
Demonstration Study
(2019) 19:268

India Microscopy center Direct ZN MGIT, LJ 6·5% Culture None 40 (27–51) 35% 2% 11% 25%
×2
Request for Application (RFA) Studies
India University-affiliated DOTS Direct ZN × MGIT 11·5% Xpert Frozen 43 (29–55) 37% 2% 15% 54%3
clinic 1 Culture Processed
Vietnam Microscopy center Direct ZN × MGIT 0% Xpert Fresh Processed 60 (52–70) 42% 0.5% 8% 59%
2 Culture
Malawi Microscopy center Direct FM × MGIT, LJ 8·7% Xpert Fresh Direct 35 (26–41) 48% 44% 16% 15%3
1 Culture
Tanzania District Hospital TB clinic Direct FM × MGIT, LJ 5·1% Culture Fresh Direct 37 (28–46) 45% 6% 29% 28%
2
Uganda District Hospital Direct FM 2x MGIT, 2x 1·3% Culture Fresh Direct 43 (30–54) 43% 48% 31% 45%
outpatient clinic ×2 LJ
Ivory Coast District Hospital Direct ZN MGIT 6·0% Xpert Fresh Processed 38 (28–44) 51% 12% 33% 13%
outpatient clinic ×2 Culture
Madagascar University-affiliated DOTS Direct FM 2x LJ --- Xpert Fresh Direct 42 (29–52) 40% --- 37% 27%
clinic ×2 Culture
Sponsored Study
Haiti (Kaku, Urban Hospital Direct FM 3x MGIT 0% Culture None --- --- --- 34% 23%
et al) outpatient clinic ×2
--- information not available
Abbreviations: TB tuberculosis, MGIT Mycobacterial Growth Indicator Tube, ZN Ziehl-Neelsen, FM fluorescence microscopy, LJ Lowenstein-Jensen
a
HIV unknown counted as negative. Reflects proportion of study population known to be HIV positive
b
Reference standard 3 used for all studies in this calculation
c
Smear microscopy results based on analysis of 1 smear only
Page 5 of 11
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 6 of 11

Fig. 2 Study quality assessment using QUADAS-2. Based on QUADAS-2 assessment, the risk of bias was judged to be unclear in 5 (38%) studies
due to patient selection issues, high for all studies due to an inadequate reference standard, and either high in 1 (8%) study or unclear in 7
studies (54%) due to flow and timing issues. In addition, applicability concerns were high in 5 (38%) studies due to patient selection issues

that observed among all adults, ranging from 97·7% (95% CI -8·0 to + 2·9) but lower sensitivity when using
(95% CI 96·1–98·6) with Standard 3 to 98·4% (95% CI Standard 3 (− 6·9, 95% CI -12·8 to − 1·0) (Table 3). Dif-
95·9–99·4) with Standard 1. There was greater hetero- ference in specificity between TB-LAMP and Xpert in
geneity in sensitivity estimates than in specificity esti- individual studies ranged from − 1 to + 3% for
mates across studies (Fig. 3). Standard 1, − 1 to + 4% for Standard 2, and − 3 to + 5%
for Standard 3 (Additional file 1: Figure S8). There
Comparison of diagnostic accuracy of TB-LAMP and Xpert was no difference in specificity of TB-LAMP and
Xpert sensitivity across individual studies ranged from Xpert regardless of reference standard used (Table 3).
65 to 97% between the three reference standards and Heterogeneity varied depending on the reference
specificity ranged from 90 to 100% (Additional file 1: standard used, with minimal heterogeneity in sensitivity
Figure S5). TB-LAMP sensitivity ranged from 48 to and specificity differences across studies with Standard 1,
93% between reference standards, and specificity and significant heterogeneity with Standard 3 (Additional
ranged from 94 to 100% (Additional file 1: Figure S6). file 1: Figure S7 and Figure S8).
The difference in sensitivity between TB-LAMP and
Xpert in individual studies ranged from − 14 to + 3% Indeterminate/invalid TB-LAMP results
for Standard 1, − 15 to + 3% for Standard 2, and − 36 The proportion of indeterminate TB-LAMP results was
to + 3% for Standard 3 (Additional file 1: Figure S7). 0% in 11 studies and 1% in two studies (Additional file 1:
TB-LAMP had similar sensitivity compared to Xpert Figure S9). There was minimal heterogeneity across
using Standard 1 (pooled sensitivity difference − 2·5% studies (I2 28%, p = 0·25). Pooled proportion of
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 7 of 11

Fig. 3 Forest plots of TB-LAMP diagnostic accuracy, best reference standard. The figures show the sensitivity and specificity of TB-LAMP in
individual studies in reference to the best available reference standard for TB-LAMP as an alternative test for smear microscopy in all patients
(Panel 3A), TB-LAMP as an alternative test for smear microscopy in HIV-positive adults (Panel 3B), and TB-LAMP as an add-on test following smear
microscopy (Panel 3C). All reference standards classify patients as having TB if ≥1 positive culture was confirmed as M. tuberculosis by speciation
testing. To be classified as not having TB, patients were required to have no positive and at least 1) two negative cultures on two different
sputum specimens (Standard 1); or 2) two negative cultures on the same or different sputum specimens (Standard 2). Visual inspection of all
three forest plots indicates considerable heterogeneity in sensitivity estimates but less heterogeneity in specificity estimates

indeterminate TB-LAMP results was < 0.1% (95% CI 0–0). no significant differences in pooled sensitivity (− 2·5,
Results were similar among HIV-infected adults; pooled 95% CI -8·0 to + 2·9) or pooled specificity (0·5, 95% CI
proportion of indeterminate TB-LAMP results in this -0·9 to + 1·8) using the most stringent reference
sub-group was < 0.1% (95% CI 0–1). standard. Finally, this review found indeterminate
TB-LAMP results were extremely uncommon (pooled pro-
Discussion portion < 0.1, 95% CI 0–0) (Additional file 1: Figure S9).
This systematic review identified 13 studies conducted Overall, these data support a potential role for TB-LAMP
in intermediate to high TB burden countries evaluating in the diagnosis of pulmonary TB in intermediate- to
the accuracy of TB-LAMP performed directly on spu- high-burden countries where smear-microscopy is still the
tum samples for diagnosis of pulmonary TB. TB-LAMP predominant mode of TB diagnosis.
had moderate sensitivity (pooled sensitivity 77·7, 95% CI The target product profile for an alternative test for
71·2–83·0) and high specificity (98·1, 95% CI 95·7–99·2) smear microscopy recommends a sensitivity of at least
when using the most stringent culture-based reference 80% and specificity of at least 98% [15]. TB-LAMP very
standard. Sensitivity was lower among HIV-infected nearly meets these criteria, although sensitivity was
adults (pooled sensitivity 63·8, 95% CI 49·0–76.4), likely below the recommended minimum of 60% for
due to a higher proportion of smear-negative TB in this smear-negative TB and 99% for smear-positive TB.
population. Among all adults, TB-LAMP would identify Nonetheless, TB-LAMP was consistently more sensitive
slightly less than half of all smear-negative TB patients than sputum smear microscopy in individual studies,
(pooled sensitivity 42·1, 95% CI 30·0–55·3) if used as an and pooled sensitivity difference was 7·1%–13·2% in
add-on test following sputum smear microscopy. Diag- favor of TB-LAMP depending on reference standard
nostic accuracy was comparable to that of Xpert, with used. Thus, it can be expected that use of TB-LAMP as
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 8 of 11

Table 2 Pooled sensitivity and specificity of TB-LAMP minimized any difference in specificity between
Reference Standard Pooled Sensitivity Pooled Specificity TB-LAMP and smear microscopy.
TB-LAMP accuracy if used as an alternative test for smear microscopy in Although more accurate than microscopy, pooled sen-
all adults sitivity of TB-LAMP is lower than has been reported for
Standard 1a 77·7 (71·2–83·0) 98·1 (95·7–99·2) Xpert (89, 95% CI 85–92] [16]. However, in
Standard 2a 76·0 (69·9–81·2) 98·0 (96·0–99·0) head-to-head comparisons, the sensitivity difference be-
a tween TB-LAMP and Xpert was not statistically signifi-
Standard 3 80·3 (70·3–87·5) 97·7 (96·1–98·7)
cant except when using the least stringent reference
TB-LAMP accuracy if used as an alternative test for smear microscopy in
HIV-positive adults
standard. Although more data is needed to confirm
whether TB-LAMP is as sensitive as Xpert, it is import-
Standard 1a N/A (< 4 studies) N/A (< 4 studies)
ant to consider TB-LAMP has a different end-user pro-
Standard 2a 63·8 (49·0–76.4) 98·8 (85·1–99·9) file. It is less costly to deploy and has fewer
a
Standard 3 73·4 (51·9–87·6) 95·0 (64·0–99·5) infrastructure requirements (e.g., stable power), but has
TB-LAMP accuracy if used as an add-on test in smear-negative adults higher training requirements due to less automation.
Standard 1a 42·1 (30·0–55·3) 98·4 (95·9–99·4) Even if confirmed to be less sensitive than Xpert,
Standard 2a 42·2 (27·9–57·9) 98·0 (96·0–99·0)
TB-LAMP may have a role at health centers that have
a personnel with adequate technical skills but insufficient
Standard 3 40·3 (27·9–54·0) 97·7 (96·1–98·6)
resources or infrastructure to deploy Xpert. In addition,
TB-LAMP accuracy in studies comparing to Xpertb the specificity of TB-LAMP was as high or higher than
Standard 1a 78·0 (66·6–86·4) 98·9 (97·4–99·6) that of Xpert, further supporting its use as an alternative
Standard 2a 74·1 (64·1–82·2) 98·8 (96·8–99·6) test for microscopy in settings without access to Xpert.
Standard 3 a
75·8 (63·2–85·0) 98·2 (96·0–99·2) There are several limitations to the evidence identified
Xpert accuracy in studies comparing to TB-LAMPb
in this review. Most significantly, an inadequate refer-
ence standard was used across all studies, likely leading
Standard 1a 81·1 (70·6–88·5) 98·2 (95·9–99·2)
to classification of patients with TB as not having TB.
Standard 2a 80·4 (73·4–85·9) 97·4 (94·9–98·7) With a better reference standard, it can be expected that
a
Standard 3 84·0 (75·6–90·0) 97·2 (94·4–98·6) some false-positive TB-LAMP results would be
a
All reference standards classify patients as having TB if ≥1 positive culture re-classified as true positives, leading to improved sensi-
was confirmed as M. tuberculosis by speciation testing. To be classified as not
having TB, patients were required to have no positive and at least 1) two
tivity and specificity. Some true negative TB-LAMP re-
negative cultures on two different sputum specimens (Standard 1); 2) two sults could also be re-classified as false negatives, leading
negative cultures on the same or different sputum specimens (Standard 2); or to lower sensitivity and specificity. Our findings suggest
3) at least one negative culture (Standard 3)
b
Data restricted to study participants who had valid results for both TB-LAMP the former is more likely, given that TB-LAMP specifi-
and Xpert and testing performed on non-frozen specimens city improved with a more stringent reference standard
Abbreviations: N/A not applicable
(i.e., when moving from Standard 3 to Standard 1). A
better reference standard may also have increased the
an alternative for sputum smear microscopy would lead sensitivity difference while further minimizing the speci-
to more TB cases being identified while keeping ficity difference observed between Xpert and TB-LAMP
false-positive results to an acceptable minimum. As dis- due to the higher number of false-positive Xpert results.
cussed further below, a better reference standard would Another key limitation is that the included studies may
have likely further increased difference in sensitivity and not accurately reflect the introduction of TB-LAMP
under programmatic conditions. Only 3 studies were
conducted at peripheral health centers and Eiken pro-
Table 3 TB-LAMP versus Xpert: Pooled Sensitivity and Specificity
differences
vided extensive training to sites included in all studies.
Further operational research is needed to characterize
Reference Pooled sensitivity Pooled specificity
standard differenceb differenceb TB-LAMP performance under typical implementation
Standard 1a -2·5 (− 8·0 to + 2·9) 0·5 (− 0·9 to + 1·8) conditions as an alternative test for microscopy at per-
ipheral health centers. Finally, the data used in this ana-
Standard 2 a
− 6·0 (− 12·1 to + 0·1) 1·0 (− 0·3 to + 2·4)
lysis was taken from studies sponsored by either a
Standard 3a − 6·9 (− 12·8 to − 1·0) 1·1 (− 0·7 to + 2·8)
non-governmental organizations (NGOs) with interest in
a
All reference standards classify patients as having TB if ≥1 positive culture the product or the manufacturer (Eiken), raising poten-
was confirmed as M. tuberculosis by speciation testing. To be classified as not
having TB, patients were required to have no positive and at least 1) two tial concerns about bias due to conflict of interest. To
negative cultures on two different sputum specimens (Standard 1); 2) two overcome this limitation, this systematic review and
negative cultures on the same or different sputum specimens (Standard 2); or
3) at least one negative culture (Standard 3)
meta-analysis were the result of independent analysis of
b
Positive difference favors TB-LAMP individual level data taken from these studies. PICO
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 9 of 11

questions and reference standards were devised by the with TB symptoms in settings where Xpert testing is not
authors in conjunction with the World Health available and where drug-resistance or HIV co-infection
Organization as part of their guideline development are not of concern [17].
process.
In summary, this systematic review supports use of Additional file
TB-LAMP as a potential alternative test for smear mi-
croscopy for diagnosis of pulmonary TB in intermediate- Additional file 1: Figure S1. Forrest plots of TB-LAMP diagnostic accuracy,
to high-burden countries, particularly in settings where additional reference standards. The figures show the sensitivity and specificity
of TB-LAMP in individual studies in reference to all additional reference
Xpert testing is not feasible. The results of this review standards not judged best available for TB-LAMP as an alternative test for
have influenced development of WHO guidelines related smear microscopy in all patients (Panel S1A and Panel S1B), TB-LAMP as an
to the use of TB-LAMP, which now conditionally rec- alternative test for smear microscopy in HIV-positive adults (Panel S1C), and
TB-LAMP as an add-on test following smear microscopy (Panel S1D and Panel
ommend the use of TB-LAMP as an alternative test for S1E). All reference standards classify patients as having TB if ≥1 positive culture
smear microscopy or as an add-on test for smear nega- was confirmed as M. tuberculosis by speciation testing. To be classified as not
tive patients, but do not recommend its use in settings having TB, patients were required to have no positive and at least 1) two
negative cultures on two different sputum specimens (Standard 1); 2) two
where molecular testing such as Xpert is readily avail- negative cultures on the same or different sputum specimens (Standard 2); or
able [17]. However, additional studies following stan- 3) at least one negative culture (Standard 3). Visual inspection of all three forest
dardized protocols and including a high-quality plots indicates considerable heterogeneity in sensitivity estimates but less
heterogeneity in specificity estimates. Figure S2. TB-LAMP as an alternative for
reference standard (liquid culture results on at least two sputum smear microscopy: Summary Receiver Operating Characteristic (SROC)
samples) are needed to better inform National TB Pro- curves. The figure shows SROC curves for TB-LAMP (green line), individual
grammes of the relative performance of TB-LAMP ver- study estimates (grey circle), pooled estimates (red square), and the 95%
confidence region for pooled estimates (yellow dotted line) when using 3
sus Xpert in programmatic settings. Cost effectiveness culture-based reference standards. All reference standards classify patients as
analysis conducted as part of WHO guideline develop- having TB if ≥1positive culture was confirmed as M. tuberculosis by speciation
ment on the use of TB-LAMP demonstrated lower per testing. To be classified as not having TB, patients were required to have no
positive and at least 1) two negative cultures on two different sputum
test cost, operational costs, budgetary costs and favor- specimens (Standard 1); 2) two negative cultures on the same or
able incremental cost effectiveness ratios compared to different sputum specimens (Standard 2); or 3) at least one negative
Xpert in a few countries [18]. As with feasibility assess- culture (Standard 3). Figure S3. TB-LAMP as an alternative test for
smear microscopy in HIV-positives: Summary Receiver Operating
ments, additional cost and cost effectiveness analyses in Characteristic (SROC) curves. The figure shows SROC curves for TB-
programmatic settings are required to further inform LAMP (green line), individual study estimates (grey circle), pooled
context-specific uptake of TB-LAMP. The evidence to estimates (red square), and the 95% confidence region for pooled
estimates (yellow dotted line) when using 2 culture-based reference
date, along with increased automation and the ability to standards (no studies qualified for Standard 1). All reference standards
identify rifampin resistance, suggests Xpert should re- classify patients as having TB if ≥1positive culture was confirmed as
main the preferred diagnostic when sufficient financial M. tuberculosis by speciation testing. To be classified as not having TB,
patients were required to have no positive and at least 1) two negative
resources and infrastructure can support its use. cultures on the same or different sputum specimens (Standard 2); or 2)
at least one negative culture (Standard 3). Figure S4. TB-LAMP as an
Conclusions add-on test following smear microscopy: Summary Receiver Operating
Characteristic (SROC) curves. The figure shows SROC curves for TB-
Although the performance of TB-LAMP has been evalu- LAMP (green line), individual study estimates (grey circle), pooled
ated in several studies worldwide with variable results, estimates (red square), and the 95% confidence region for pooled
we report the first standardized evaluation of the results estimates (yellow dotted line) when using 3 culture-based reference
standards. All reference standards classify patients as having TB if
of these individual studies to inform policy guidance. ≥1positive culture was confirmed as M. tuberculosis by speciation
The findings of our systematic review and meta-analysis testing. To be classified as not having TB, patients were required to
show that TB-LAMP has the potential to be a useful have no positive and at least 1) two negative cultures on two different
sputum specimens (Standard 1); 2) two negative cultures on the same
diagnostic test for pulmonary TB, but that additional or different sputum specimens (Standard 2); or 3) at least one negative
high quality studies in programmatic conditions should culture (Standard 3). Figure S5. TB-LAMP vs. Xpert MTB/RIF: Forest plots
be done to strengthen the case for its use. TB-LAMP of Xpert MTB/RIF diagnostic accuracy. The figure shows forest plots of
Xpert MTB/RIF sensitivity and specificity in reference to 3 culture-based
performed better than sputum smear microscopy (more reference standards for individual studies. All reference standards
sensitive and as specific) in the diagnosis of pulmonary classify patients as having TB if ≥1positive culture was confirmed as M.
tuberculosis and performed similar to Xpert (similar tuberculosis by speciation testing. To be classified as not having TB,
patients were required to have no positive and at least 1) two negative
sensitivity and specificity). The results of this study pro- cultures on two different sputum specimens (Standard 1); 2) two
vided the basis for the WHO’s guidelines on the use of negative cultures on the same or different sputum specimens (Standard
TB-LAMP for the diagnosis of pulmonary TB, which 2); or 3) at least one negative culture (Standard 3). Figure S6. TB-LAMP
vs. Xpert MTB/RIF: Forest plots of TB-LAMP diagnostic accuracy. The
recommend that TB-LAMP can be used as an alterna- figure shows forest plots of TB-LAMP sensitivity and specificity in
tive for microscopy for the diagnosis of pulmonary TB reference to 3 culture-based reference standards for individual studies.
in adults, and can be considered as an add-on test to mi- All reference standards classify patients as having TB if ≥1positive
culture was confirmed as M. tuberculosis by speciation testing. To be
croscopy particularly in sputum smear-negative adults
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 10 of 11

classified as not having TB, patients were required to have no positive and at Authors’ contributions
least 1) two negative cultures on two different sputum specimens (Standard PS, KF, and LS conducted the literature search. All authors contributed to the
1); 2) two negative cultures on the same or different sputum specimens design of the study. PS and KF developed the figures for the manuscript.
(Standard 2); or 3) at least one negative culture (Standard 3). Figure S7. TB- Any additional data collection from Eiken or other sources was conducted
LAMP vs. Xpert MTB/Rif: Forest plots of sensitivity difference. The figure shows by PS, KF, LS, and CG. PS KF LS CG collaborated on data analysis. Data were
forest plots of the sensitivity difference between TB-LAMP and Xpert MTB/Rif® interpreted by PS, KF, LS, and AC. PS and KF wrote the initial version of the
for individual studies. The sensitivity of both tests was calculated in reference manuscript and all authors contributed to review and editing of the final
to 3 culture-based reference standards. All reference standards classify patients version. The corresponding author had full access to all the data and had
as having TB if ≥1positive culture was confirmed as M. tuberculosis final responsibility for the decision to submit for publication. All authors read
by speciation testing. To be classified as not having TB, patients were and approved the final manuscript.
required to have no positive and at least 1) two negative cultures on
two different sputum specimens (Standard 1); 2) two negative Ethics approval and consent to participate
cultures on the same or different sputum specimens; or 3) at least Ethics approval for this study was waived (University of California, San
one negative culture (Standard 3). Visual inspection of forest plots Francisco Internal Review Board) as it involved analysis only of previously
and statistical testing indicate minimal heterogeneity with Standard 1 collected de-identified data received by the authors from Eiken and from in-
(I2 0%, p = 0.41), some heterogeneity with Standard 2 (I2 34%, p = dividual study sites.
0.18), and significant heterogeneity with Standard 3 (I2 55%, p = 0.03).
Figure S8. TB-LAMP vs. Xpert MTB/Rif®: Forest plots of specificity Consent for publication
difference. The figure shows forest plots of the specificity difference Not applicable.
between TB-LAMP and Xpert MTB/Rif® for individual studies. The
specificity of both tests was calculated in reference to 3 culture- Competing interests
based reference standards. All reference standards classify patients as The authors declare they have no competing interests.
having TB if ≥1positive culture was confirmed as M. tuberculosis by
speciation testing. To be classified as not having TB, patients were re- Publisher’s Note
quired to have no positive and at least 1) two negative cultures on Springer Nature remains neutral with regard to jurisdictional claims in
two different sputum specimens (Standard 1); 2) two negative published maps and institutional affiliations.
cultures on the same or different sputum specimens (Standard 2); or
3) at least one negative culture (Standard 3). Visual inspection of forest plots Author details
and statistical testing indicate minimal heterogeneity with Standard 1 1
Division of Pulmonary and Critical Care Medicine, University of California
(I2 28%, p = 0.25) and Standard 2 (I2 37%, p = 0.16), but significant San Francisco and Zuckerberg San Francisco General Hospital, 1001 Potrero
heterogeneity with Standard 3 (I2 72%, p = 0.001). Figure S9. Avenue, 5K1, San Francisco, CA 94110, USA. 2Curry International Tuberculosis
Proportion of indeterminate TB-LAMP results. The figure shows a Center, University of California San Francisco, San Francisco, CA, USA.
forest plot of the proportion of indeterminate TB-LAMP results 3
Division of Infectious Diseases, Oregon Health & Science University, 3188 SW
among all adults for individual studies. Visual inspection of forest Sam Jackson Park Road, Mail Code: L457, Portland, OR 97239, USA. 4Faculty
plots and statistical testing indicate minimal heterogeneity (I2 28%, p of Epidemiology and Population Health, London School of Hygiene and
= 0.25). Table S1. Patients included for analysis. Table S2. Signaling Tropical Medicine, Room G30, Keppel Street, London WC1E 7HT, UK.
questions for QUADAS-2 domains. Table S3. TB-LAMP as an alternative test
for smear microscopy: Exploration of heterogeneity. (DOCX 770 kb) Received: 21 September 2017 Accepted: 5 March 2019

Abbreviations References
AFB: Acid fast bacilli; CI: Confidence interval; DEMO: Demonstration study; 1. World Health Organization: Automated real-time nucleic acid amplicfication
DNA: Deoxyribonucleic acid; EVAL: Evaluation study; FIND: Foundation for technology for rapid and simultaneous detection of tuberculosis and
Innovative Diagnostics; FM: Fluorescence microscopy; HIV: Human rifampicin resistance: Xpert MTB/RIF assay for the diagnosis of pulmonary
immunodeficiency virus; HSROC: Hierarchical summary reciever operating and extrapulmonary TB in adults and children. In.; 2013.
characteristic; LAMP: Loop-mediated isothermal amplification; 2. Notomi T, Okayama H, Masubuchi H, Yonekawa T, Watanabe K, Amino N,
MTB: Mycobacterium tuberculosis; NTM: Non-tuberculous mycobacteria; Hase T. Loop-mediated isothermal amplification of DNA. Nucleic Acids Res.
QUADAS: Quality Assessment of Studies of Diagnostic Accuracy for 2000;28(12):E63.
Systematic Reviews; RFA: Request for Applications; RIF: Rifampin; 3. Iwamoto T, Sonobe T, Hayashi K. Loop-mediated isothermal amplification
TB: Tuberculosis; UV: Ultraviolet; WHO: World Health Organization; ZN: Ziehl- for direct detection of mycobacterium tuberculosis complex, M. Avium, and
Neelsen M. Intracellulare in sputum samples. J Clin Microbiol. 2003;41(6):2616–22.
4. Sethi S, Singh S, Dhatwalia SK, Yadav R, Mewara A, Singh M, Tewari R,
Sharma M. Evaluation of in-house loop-mediated isothermal amplification
Acknowledgements (LAMP) assay for rapid diagnosis of M. Tuberculosis in pulmonary
Not applicable. specimens. J Clin Lab Anal. 2013;27(4):272–6.
5. Kaewphinit T, Arunrut N, Kiatpathomchai W, Santiwatanakul S, Jaratsing P,
Funding Chansiri K. Detection of mycobacterium tuberculosis by using loop-
Funding for this study was provided by the Global TB Programme of the mediated isothermal amplification combined with a lateral flow dipstick in
World Health Organization. The funder of the study (WHO) established the clinical samples. Biomed Res Int. 2013;2013:926230.
review questions to be addressed and reviewed the study protocol. The 6. Aryan E, Makvandi M, Farajzadeh A, Huygen K, Bifani P, Mousavi SL, Fateh A,
funder had no role in data collection, data analysis, data interpretation, or Jelodar A, Gouya MM, Romano M. A novel and more sensitive loop-
writing of the manuscript. The funder did incorporate a report of these mediated isothermal amplification assay targeting IS6110 for detection of
results into their published guidelines on the use of TB-LAMP. mycobacterium tuberculosis complex. Microbiol Res. 2010;165(3):211–20.
7. Mitarai S, Okumura M, Toyota E, Yoshiyama T, Aono A, Sejimo A, Azuma Y,
Sugahara K, Nagasawa T, Nagayama N, et al. Evaluation of a simple loop-
Availability of data and materials mediated isothermal amplification test kit for the diagnosis of tuberculosis.
The data that support the findings of this study are available from Eiken, but Int J Tuberc Lung Dis. 2011;15(9):1211–7 i.
restrictions apply to the availability of these data, which were used at the 8. Higgins J, Green S, Cochrane Collaboration., Wiley InterScience (Online
discretion of Eiken for the current study, and so are not publicly available. service). Cochrane handbook for systematic reviews of interventions. In:
Data are however available from the authors upon reasonable request and Cochrane book series. Chichester, England; Hoboken, NJ: Wiley-Blackwell;
with permission of Eiken. 2008. 1 online resource (xxi, 649 p.).
Shete et al. BMC Infectious Diseases (2019) 19:268 Page 11 of 11

9. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB,
Leeflang MM, Sterne JA, Bossuyt PM, Group Q. QUADAS-2: a revised tool for
the quality assessment of diagnostic accuracy studies. Ann Intern Med.
2011;155(8):529–36.
10. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis.
Stat Med. 2002;21(11):1539–58.
11. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in
meta-analyses. BMJ. 2003;327(7414):557–60.
12. Gray CM, Katamba A, Narang P, Giraldo J, Zamudio C, Joloba M, Narang R,
Paramasivan CN, Hillemann D, Nabeta P, et al. Feasibility and operational
performance of tuberculosis detection by loop-mediated isothermal
amplification platform in decentralized settings: results from a multicenter
study. J Clin Microbiol. 2016;54(8):1984–91.
13. Nliwasa M, MacPherson P, Chisala P, Kamdolozi M, Khundi M, Kaswaswa K,
Mwapasa M, Msefula C, Sohn H, Flach C, et al. The sensitivity and specificity of
loop-mediated isothermal amplification (LAMP) assay for tuberculosis diagnosis
in adults with chronic cough in Malawi. PLoS One. 2016;11(5):e0155101.
14. Kaku T, Minamoto F, D'Meza R, Morose W, Boncy J, Bijou J, Geffrard H,
Yoshida M, Mori T. Accuracy of LAMP-TB method for diagnosing
tuberculosis in Haiti. Jpn J Infect Dis. 2016;69(6):488–92.
15. World Health Organization. High-priority target product profiles for new
tuberculosis diagnostics: report of a consensus meeting. Geneva: World
Health Organization; 2014.
16. Steingart KR, Schiller I, Horne DJ, Pai M, Boehme CC, Dendukuri N. Xpert(R)
MTB/RIF assay for pulmonary tuberculosis and rifampicin resistance in
adults. Cochrane Database Syst Rev. 2014. https://doi.org/10.1002/14651858.
CD009593.pub3.
17. World Health Organization. The use of loop-mediated isothermal
amplification (TB-LAMP) for the diagnosis of pulmonary tuberculosis. Policy
guidance. Geneva: World Health Organization; 2016.
18. Sohn H. Cost, affordability, and cost-effectiveness of TB-LAMP assay. In:
Report to WHO Guideline Development Group Meeting on TB-LAMP Assay.
Edn. Geneva: World Health Organization; 2016.