It isn’t
easy to find a test that is as fast and as sensitive as Xpert MTB/Rif. But
there could be one candidate that approaches MTB/RIF in sensitivity while being
simple enough to use at the point of care in even the most remote low-resourced
sites — a recently developed lateral flow version of the LAM urine antigen
test.
Liparabinomannan (LAM) is a glycolipid
component of the TB outer cell wall. When it was shown to be detectable in
urine in certain animal models interest grew in LAM as a potential diagnostic
target in human TB. It was an attractive target because it was a direct product
of the mycobacteria (rather than an immune response to it — antibodies have not
proven to be good diagnostic targets for TB), and because urine is easier
to handle, without the infection control concerns that come with handling
sputum or blood. Furthermore, urine is suitable for use in inexpensive
point-of-care lateral flow strip tests.
NAM/HATIP has written about LAM several
times over the years — initially excited about great results from a study in Tanzania — but then two studies elsewhere in
sub-Saharan Africa could not reproduce the
results. These evaluated an ELISA version of the test for detection of LAM in
urine – that assay is a direct sandwich immunoassay that uses polyclonal
antibodies. However, while the ELISA didn’t look as useful in the general
population of people with TB, in adults living with advanced HIV disease the
sensitivity of Determine TB-LAM for culture-positive TB was approximately 56%,
with specificity of 91-95%.7 This means that it is an interesting rule-in test for TB. But to be
cost-effective, specimens had to be batched and run on the ELISA test, which
delayed diagnosis. This is not the case with the new lateral flow version,
which is being marketed as a point-of-care test that can be performed on the urine sample
of any HIV
positive TB suspect who needs it, with results that should be available about
30 minutes to an hour after the urine is applied to the test. It costs around
US $2.50 - $3.50 per test.
Dr Jonny
Peter of the University
of Cape Town presented
some preliminary data on the new test at the South African AIDS conference last
June.8
The hand-held test has a urine loading platform on one side, where a 60µl urine
specimen is applied. Unlike the ELISA, no urine sample processing is required. A spot of urine is placed on to the base of the strip; the detection system uses polyclonal antibodies and colloidal gold; and the reporting scale is semi-quantitative. No band is a negative result and any bands present are then graded in intensity from 1+ to 5+. But as per the manufacturers recommendations, the presence of any band is read as a positive test.
To
evaluate the assay, 335 hospitalised people with suspected TB/HIV coinfection,
and 88 non-TB control cases (also HIV-positive) were recruited at four secondary hospitals in
Cape Town. 87% of
the participants were HIV-infected, with a median interquartile CD4 cell count
range of 115 (54-243).
146 were culture-positive TB cases. Smear microscopy in this study was 51%
sensitive, indicating just how sensitive this test can be in some
circumstances.
As for
the performance of Determine TB-LAM using the manufacturer’s cut-point of grade 1, inter-observer agreement was moderate
(κ=0.51), the sensitivity was 68% (74% in CD4 cell counts of 200 or below), but
the specificity was on the low side at 90%. However, using a cut-point of grade 2 for rapid ‘rule-in’, inter-observer agreement was good (κ=0.89), the sensitivity
was 50% (59% in CD4≤200 cells) and the specificity was 97%.
“As for
its clinical utility, we can say that point-of-care LAM is a rapid, inexpensive
‘rule-in’ of TB in hospitalized, HIV-coinfected patients for commencement of
early TB treatment,” said Dr Peter.
In
addition, the sensitivity of point-of-care LAM in smear-negative, culture-positive TB was
45%. When the Determine LAM test was combined with smear microscopy (which
seemed unusually sensitive in this study), the combined
sensitivity was 70% for culture-positive TB.
Dr Peter
cautioned that LAM should not be seen as a replacement for microbiological TB
diagnosis, and DST where possible — but as has already been mentioned, culture
and DST are not available everywhere.
Determine TB-LAM compared with Xpert MTB/RIF
A second
study, conducted by Dr Stephen Lawn and colleagues at the Desmond Tutu HIV
Centre, enrolled 602 adults referred for TB screening at a community-based ART
clinic in Gugulethu township near Cape
Town.9
Sputum samples were obtained for fluorescence microscopy, automated liquid
culture and Xpert MTB/RIF assays, and urine samples for the Clearview TB-ELISA and
the Determine TB-LAM test.
Of those
enrolled, 542 subjects provided one or more sputum samples and 94 of the
patients had culture-positive tuberculosis (prevalence 17.4%, 95% CI 14.2–20.8).
516 patients had complete results available for all the tests (including 85 of
the culture-positive patients). These were relatively advanced patients, with a
median CD4 cell count of 160.
It is
possible that the reference scale card was modified since Dr Peter’s study, but
in this study, any result above negative was treated as a positive result.
The inter-reader agreement was very high between the two readers (κ=0·97) —
although it was the policy in this study that if there was too great a
discrepancy between the readers, they would wait another 25-35 minutes and come
back to read the test results again, so this may have increased agreement at
the end of the day — and between the test strips and TB-ELISA (κ=0·84).
Lawn et
al stratified the results for Determine TB-LAM by CD4 count, and noted that
test was far more sensitive for TB among those with the lowest CD4 cell counts —
the very group of patients for whom a rapid diagnosis is most urgent — 66·7%
positive (95% CI 41·0–86·7) below 50 cells, 51·7% (32·5–70·6) at <100 cells, and 39·0%
(26·5–52·6) at <200 cells; specificity was greater than 98% for all strata.
Similar
to the finding in the Peter study,
combining the Determine TB-LAM with smear microscopy increased sensitivity to 72.2%
(95% CI 46.5–90.3) at CD4 counts less than 50 cells, 65.5% (45.7–82.1) at less
than 100 cells, and 52.5% (39.1–65.7) at less than 200 cells. The gain in
sensitivity achieved by combining the methods was more pronounced in patients
with higher CD4 cell counts.
Then the researchers compared these results to the sensitivity and specificity
of the Xpert MTB/RIF when testing a single sputum sample. When all the CD4
counts were included, Xpert was more sensitive than Determine TB-LAM test
strips and microscopy combined. But in the patients with less than 100
and less than 50 CD4 cells, the sensitivity of Determine TB-LAM and smear
microscopy combined did not differ significantly from that of the Xpert MTB/RIF
assay in patients from those strata. Assuming that further research shows that
the test works as well in other African settings, the Determine TB-LAM may
provide an affordable alternative for those settings that can’t afford the capital
outlays for GeneXpert equipment, particularly when combined with microscopy.
Determine TB-LAM may
provide an affordable alternative for those settings that can’t afford the capital
outlays for GeneXpert equipment, particularly when combined with microscopy.
In
addition, there was a small increase in sensitivity when results from both
Xpert MTB/RIF and Determine TB-LAM were combined — and the authors noted
that combining the two might be an alternative solution to the expensive
strategy of performing two GeneXperts to diagnose smear-negative TB in PLHIV.
The
authors noted that the negative predictive value for Determine TB-LAM was not
high enough to rule out TB when there is a negative result, and wrote: “The
assay should
be restricted for use as a test for tuberculosis in patients with advanced
immunodeficiency. These are the patients in whom tuberculosis diagnosis is so
challenging such that, in the absence of suitable diagnostic assays, empirical
treatment has been suggested as a strategy to reduce the high mortality of
patients with very low CD4 cell counts in settings with the highest disease
burden. However, with the development of
this simple point-of-care assay, such a strategy might prove unnecessary. Studies
of the use of this assay and its effects on clinical outcomes in such patient
groups are now needed.”
In
addition, studies are urgently needed to reproduce these findings in other
settings and to determine whether the test can be used in isolation for
reliable rapid tuberculosis diagnosis in this population, or whether the
results have to be confirmed with another lab test before they can be acted on
clinically (to inform the decision to treat).
At the
2011 World Lung Health conference in Lille,
only two oral presentations made any significant mention of the LAM lateral
flow test. During the New Diagnostics Working Group Annual Meeting, Professor
Ruth McNerney of the London School of
Hygiene & Tropical Medicine gave a talk on the prospects of a true point-of-care TB test
in the near future, and described the Determine TB LAM test as having many of
the qualities of a point-of-care test, except perhaps its narrow applicability to people
with advanced HIV disease alone — and the fact that “it is a ‘rule-in’ test —
not a ‘rule-out’ test.”10
The other
presentation was from a Cambodian study evaluating Xpert MTB/Rif
and LAM in 829 children, 121 of whom had culture-proven TB. Xpert was negative
in all the induced sputum and stool samples but was able to detect TB in some gastric aspirates.11
LAM on the other hand was positive in 12 of the children with culture-positive
TB (10%) but was positive on 28 cases that did not have TB (specificity 84%). “Urine
LAM showed poor test characteristics,” the researchers concluded.
Other
than that, only two posters reported on LAM, but not the lateral flow test. One
poster described a systematic review on diagnostic accuracy of mycobacterial antigen
detection tests for pulmonary and extrapulmonary TB.12
This mostly concerned LAM, since it is the only TB antigen to have been investigated in a large number of studies, and included a meta-analysis of studies
(involving a number of test formats, 60% of which used urine as the specimen
tested) published up to August 2010: “Compared with HIV-uninfected patients [4
studies, sensitivity 14% (95% CI 4–38), specificity 97% (95% CI 86–100)],
urinary LAM detection in HIV-infected patients (5 studies) yielded higher
sensitivity for TB [47% (95% CI 26–68)] and similar specificity [96% (95% CI
81–100).” Meanwhile the ability to diagnose extrapulmonary TB by detecting
antigens in cerebrospinal fluid (CSF), serum, biopsy, pleural fluid, urine, and
lymph node aspirates has been extremely variable. “No current test is
sufficiently accurate for active TB diagnosis. Considering that such tests can
be translated into rapid and inexpensive point-of-care tests, research to
improve performance of these tests is imperative,” the authors wrote.
LAM as a predictor for TB-IRIS
The
other poster involved the Clearview LAM ELISA in people with advanced HIV disease — but asked
a rather novel question: is LAM detectability in people living with HIV who are
on treatment for active TB disease associated with the development of TB Immune
Reconstitution Inflammatory Syndrome (TB-IRIS) when they go onto antiretroviral
therapy — and could it be used to predict it?13
The
study was a prospective observational cohort at Mulago
Hospital, in Kampala, Uganda,
which was followed for about 11 months. Cases were defined as those patients who
developed TB IRIS during the first 3 months of HAART (the definition from the
International Network for the Study of HIV-associated IRIS (INSHI)), while controls
were patients who remained TB IRIS-free during the follow up period.
There
were 26 IRIS cases and 66 controls included in the analysis. The median time to
development of IRIS was 14 days. 81% of the cases had a positive LAM test
compared to 48% among the controls. In the univariate analysis a positive
urinary LAM test pre-HAART had an odds ratio of 4.5 [95% CI:1.5–13.3], P = 0.007) while a baseline CD4 cell T-cell count
< 50 cells had an odds ratio of 21 [95%
CI:2.6–168.8], P = 0.004) for an
increased risk of TB IRIS. But only the baseline CD4 cell count was predictive
of IRIS (P < 0.001).
Overall,
LAM ELISA’s sensitivity for an IRIS diagnosis was 80.8% (60.6–93.4) with a low
specificity of 51.6% (95%CI: 38.7–64.2). The researchers concluded that, “a
positive LAM test was associated with TB IRIS. In the absence of CD4 T-cell
count, LAM detection in a point-of-care format could be useful to detect
patients at high risk of TB IRIS. Further testing with lateral flow point-of-care
format is warranted.”
In other
words, it sounds as though a point-of-care LAM test could be a surrogate marker test
for having a very low CD4 cell count in people with HIV/TB.
The low
specificity for IRIS isn’t that surprising because all of these people had
advanced HIV/TB. However, the idea that LAM antigen is more likely to be
detected by the test because it is more likely to be shed in the most advanced
patients or those with the poorest immune control is interesting because it
might also be shed in greater quantities in the patients most at risk — and the
Determine TB LAM might be able to distinguish between a lot of antigen versus
just a little with its cut points of 1-5. By using a higher cut-off point, it
might indeed be possible to identify those at higher risk and exclude those who are not. It is worth further study.
But it
must be said that, for such a huge conference on tuberculosis, there was
noticeably little attention given to LAM, which could be for two reasons. Either the
hardcore TB experts don’t trust it, and think it won’t really pan out in future
studies (which is quite possible), or they are disinterested because it is just
for patients with the most advanced HIV disease — and think it probably won’t make a
very important contribution to TB control. If that is the case, HIV programme
people and activists may have to take the lead in making sure that Determine TB
LAM is adequately evaluated and if warranted, integrated into the clinical
management algorithms for TB/HIV.
LAM in HIV-positive populations
Indeed, there appeared to be more interest in the test on the final day of an
HIV conference, CROI, where Dr Susan Dorman presented the interim results of a
multicentre study of the lateral flow test for diagnosis of TB in adults with
HIV infection.14
The main
study objectives were to estimate the sensitivity and specificity of the Determine
LAM lateral flow assay in adult HIV-positive TB suspects; and to compare the
performance characteristics of the lateral flow point-of-care test versus that
of the existing in-laboratory, more complicated, ELISA version. They are also
trying to determine operating characteristics of the LAM lateral flow assay
(test failures, between-reader variability), and whether there are any clinical
characteristics associated with a positive LAM test.
The study
was mainly cross-sectional with limited longitudinal follow-up in the two
settings, each of which included an in-patient as well as an out-patient
setting: Mulago Hospital (the same team as the TB-IRIS study) and nearby
outpatient clinics in Kampala, Uganda, and at GF Jooste Hospital in Cape Town
and two outpatient clinics in Khayelitsha township.
The
eventual target sample size is 1000 participants with HIV who are suspected to
have active TB (at last one symptom from the WHO TB symptom screen). After enrolment participants underwent a brief interview; two
sputa were collected for smear, MGIT, mycobacterial culture, and LJ solid
culture. All participants had blood drawn for mycobacterial blood culture and CD4 cell
counts, had a chest x-ray and urine was obtained for the point-of-care lateral
flow testing as well as batched for near-term ELISA testing. During the study a
subset of participants, namely those who had any positive LAM test at
enrolment, but no positive culture for TB, underwent an in-person follow-up two
months later at which point the baseline investigations were repeated. Everyone
else had a record review at two months.
For the purpose of the interim analysis, Dr
Dorman presented only the results at baseline — so anybody with a positive
mycobacterial culture for TB was defined as a TB case – all others were not TB
cases.
561 participants were enrolled as of the
time of this analysis and 409 were eligible for analysis, strictly based on
their date of enrolment. About 60% were women; largely young adults in their 30s
and 40s. Approximately one-third were on ART at enrolment. There was a significant
difference between Uganda
and South Africa – Ugandan patients
had more advanced HIV disease, with a median CD4 cell count of 78 and approximately
three-quarters were enrolled in hospital, whereas in South Africa, the median CD4 cell
count was 234, and fewer than half were enrolled in hospital.
Approximately one-third had a TB culture of
sputum and/or blood that was positive. Strikingly in Uganda 16% of participants had a
blood culture that was positive for TB. The sensitivity of sputum smear
microscopy among TB cases was about 43%. Dr Dorman displayed the distribution of LAM
lateral flow results by band intensity (cut point): 68% of tests had no band;
12% of tests had a band of intensity 1+; and 21% of tests had a band intensity
greater than 1+.
But just like Dr Peter’s study, this study
found problems when any band was considered positive. The sensitivity was about
62%, but the specificity was about 78%. However, if one were to use more stringent
criteria for positivity and increase the cut point to 2+ , the sensitivity
decreased to about 45%, and specificity increased to just over 90%. If one were
to increase the cut point further, the sensitivity falls off fairly abruptly
without much gain in specificity.
Results of conventional
mycobacteriology among analysis-eligible participants
|
OVERALL
N = 409
N (%)
|
UGANDA
N = 211
N (%)
|
CAPE TOWN
N = 198
N (%)
|
MTB in sputum and/or blood (“TB case”)
|
125
(31%)
|
80 (38%)
|
45 (23%)
|
MTB
in sputum
|
121 (30%)
|
78
(37%)
|
43
(22%)
|
MTB
in blood
|
43 (11%)
|
33
(16%)
|
10
(5%)
|
Sputum
smear+ among TB cases
|
54/125 (43%)
|
35/80
(44%)
|
19/45
(42%)
|
Dr Dorman said that in subsequent analyses,
if a positivity cut point of 2+ was used and a 1+ band was considered negative,
the performance of the lateral flow point-of-care assay was strikingly similar
to that of the existing ELISA laboratory assay.
Just like Dr Lawn’s study, they were able
to show that lateral flow assay performance varied by CD4 count.
“The sensitivity of the assay is strikingly
higher in those with lower CD4 counts than in those with higher CD4 counts. In
terms of specificity – the specificity fell off a little bit as the CD4 count
was reduced. I think there are some reasons for this, that I hope we will be
able to discern once a clinical definition of TB is brought into the picture in
the final analysis,” she said.
Lateral flow sensitivity in TB patients was
similarly high if their TB was cultured from blood cultures or their CD4 cells
were low. The sensitivity was 79% in those with mycobacteraemia, (p < 0.001) and
24% in patients without mycobacteraemia. (p < 0.001), while the sensitivity
was 66.7% in those with CD4 < 100 (p < 0.001), and 20.3% among those with
CD4 > 100 (p < 0.001).
The last patient completed follow-up in
February. All cultures were expected to be mature sometime in April, and analysis should be completed by June 2012.
Dr Dorman believes that the study has shown
that using a cut-off of cut-point of 2+ may optimise sensitivity and
specificity, which seems to be supported by the fact that when using that cut point, it
has accuracy similar to the ELISA format.
She also suspects the loss of specificity
at the lowest CD4 cell counts is due to a greater frequency of culture-negative
TB in very immune-deficient individuals.
“A planned incorporation of a clinical TB
definition may result in higher observed LAM test specificities if cultures
were ‘falsely negative’, as expected in some participants,” she said. This may
be added by documenting some clinical outcomes associated with LAM detection at
low CD4 cell strata. She noted that the mortality was strikingly high,
especially in Uganda,
and that they plan to analyse the association between LAM results and
mortality, before offering some provisional conclusions.
“This
POC test performed best in the subset of patients that are most challenging to
diagnose, and have the highest mortality,” she said. Dr Dorman added that
she thinks proof of principle has been established for LAM detection, and that a
substantial research and development effort should go into refining how LAM is
detected — in other words, could some of the problems with sensitivity and
specificity be resolved by developing a better test for it?
Finally, she said she believes that
clinical trials are warranted in looking at how use of the lateral flow test in
clinical management will affect patient outcomes, “especially in ill hospitalised
patients with low CD4 counts and in pre-ART settings – as Steve Lawn and his
group have nicely explored,” she said.