Skip to main content

Full text of "Challenges in the Development of an Immunochromatographic Interferon-Gamma Test for Diagnosis of Pleural Tuberculosis."

See other formats

OPEN 3 ACCESS Freely available online 


Challenges in the Development of an 
Immunochromatographic Interferon-Gamma Test for 
Diagnosis of Pleural Tuberculosis 

Claudia M. Denkinger^'^*, Yatiraj Kalantri^, Samuel G. Schumacher^, Joy S. Michael'*, Deepa Shankar^, 
Arvind Saxena^, Natarajan Sriram^ Thangakunam Balamugesh^ Robert LuoS Nira R. Pollock^ ^ 
Madhukar Pai^, Devasahayam J. Christopher^ 

1 Division of Infectious Diseases, Betli Israel Deaconess Medical Center, Boston, Massachusetts, United States of America, 2 McGill International TB Centre & 
Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada, 3 Tulip Diagnostics, Goa, India, 
4 Department of Microbiology, Christian Medical College, Vellore, India, 5 Department of Pulmonary Medicine, Christian Medical College, Vellore, India, 
6 Department of Pathology, Stanford University, Stanford, California, United States of America, 7 Department of Laboratory Medicine, Boston Children's 
Hospital, Boston, Massachusetts, United States of America 


Existing diagnostic tests for pleural tuberculosis (TB) have inadequate accuracy and/or turnaround time. Interferon- 
gamma (IFNg) has been identified in many studies as a biomarker for pleural TB. Our objective was to develop a 
lateral flow, immunochromatographic test (ICT) based on this biomarker and to evaluate the test in a clinical cohort. 
Because IFNg is commonly present in non-TB pleural effusions in low amounts, a diagnostic IFNg-threshold was first 
defined with an enzyme-linked immunosorbent assay (ELISA) for IFNg in samples from 38 patients with a confirmed 
clinical diagnosis (cut-off of 300pg/ml; 94% sensitivity and 93% specificity). The ICT was then designed; however, its 
achievable limit of detection (5000pg/ml) was over 10-fold higher than that of the ELISA. After several iterations in 
development, the prototype ICT assay for IFNg had a sensitivity of 69% (95% confidence interval (CI): 50-83) and a 
specificity of 94% (95% CI: 81-99%) compared to ELISA on frozen samples. Evaluation of the prototype in a 
prospective clinical cohort (72 patients) on fresh pleural fluid samples, in comparison to a composite reference 
standard (including histopathological and microbiologic test results), showed that the prototype had 65% sensitivity 
(95% CI: 44-83) and 89% specificity (95% CI: 74-97). Discordant results were observed in 15% of samples if testing 
was repeated after one freezing and thawing step. Inter-rater variability was limited (3%; lout of 32). In conclusion, 
despite an iterative development and optimization process, the performance of the IFNg ICT remained lower than 
what could be expected from the published literature on IFNg as a biomarker in pleural fluid. Further improvements in 
the limit of detection of an ICT for IFNg, and possibly combination of IFNg with other biomarkers such as adenosine 
deaminase, are necessary for such a test to be of value in the evaluation of pleural tuberculosis. 

Citation: Denl<inger CM, Kalantri Y, Schumacher SG, Michael JS, Shankar D, et al. (2013) Challenges in the Development of an Immunochromatographic 
Interferon-Gamma Test for Diagnosis of Pleural Tuberculosis. PLoS ONE 8(12): e85447. doi:10.1371/journal.pone.0085447 

Editor: Delia Goletti, National Institute for Infectious Diseases (L. Spallanzani), Italy 

Received October 9, 2013; Accepted December 5, 2013; Published December 23, 2013 

Copyright: © 2013 Denkinger et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits 
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. 

Funding: This work was supported by Grand Challenges Canada award (0026-01-04-01-01 ), and by a grant from the Canadian Institutes of Health 
Research (grants MOP-89918). MP is supported by the European and Developing Countries Clinical Trials Partnership (EDCTP - TBNEAT grant) and the 
Fonds de recherche du Quebec - Sante (PROS). CMD is supported by a Richard Tomlinson Fellowship at McGill University and a fellowship of the 
Burroughs-Wellcome Fund from the American Society of Tropical Medicine and Hygiene. SGS is supported by the Quebec Respiratory Health Training 
Program (QRHTP). NRP is supported by NIH K23 AI074638-01 A2. The funders had no role in the analysis of data and decision to publish. 

Competing interests: Tulip Group, Goa, India, received grant funding from the Grand Challenges Canada grant to develop the prototype assay. MP 
serves as a consultant for the Bill & Melinda Gates Foundation. BMGF had no involvement in this manuscript. MP also serves as a section editor to PLOS 
ONE. Please note, that Yatiraj Kalantri, Arvind Saxena and Natarajan Sriram are employees of Tulip Diagnostics. This does not alter the authors' 
adherence to all the PLOS ONE policies on sharing data and materials. 

* Email; 


Extrapulmonary TB (EPTB) accounts for approximately 25% 
of all TB, and poses major diagnostic challenges. Pleural TB is 
the second most common manifestation of EPTB after lymph 
node TB [1,2]. Existing diagnostic tests have Inadequate 

accuracy and turnaround time, and require special expertise for 
sample acquisition and Interpretation of results. A pleural fluid 
aspiration rarely yields a definite diagnosis. A biopsy of the 
pleural tissue for histology and culture is considered the 
diagnostic gold standard, but may still be falsely negative In 
10% to 20% of cases [3,4]. Nucleic acid amplification tests 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 

(NAATs) for evaluation of TB in pieural effusions appear to 
have high specificity (98%) but relatively low sensitivity (62%) 
[5]. Xpert MTB/RIF (Cepheid Inc., Sunnyvale, CA), a recently 
developed NAAT, similarly had low sensitivity (25-50%) across 
a number of studies on pleural fluid (and one study on pleural 
biopsy), with consistently high specificity [6-8]. 

Biomarkers present in pleural fluid have been evaluated for 
the diagnosis of pleural TB. More than 100 studies, 
summarized in meta-analyses, indicate that adenosine 
deaminase (ADA) and interferon-gamma (IFNg) are accurate 
biomarkers of pleural TB [9,10]. ADA is released by activated 
lymphocytes and macrophages and Is a non-specific marker of 
Inflammation. Meta-analyses show that ADA (measured by its 
enzymatic activity, at varying cut-points) in pleural fluid 
specimens is 88-100% sensitive and 83-97% specific for the 
diagnosis of pleural TB [3,10-15]. Measurement with 
immunologic methods (e.g. enzyme-linked immunosorbent 
assay (ELISA) [16]) has not been explored for the detection of 
ADA in pleural fluid. 

IFNg is a soluble cytokine that is secreted by Thi cells, 
cytotoxic T cells and NK cells and has pro-inflammatory 
properties [17]. IVIeta-analyses show that free, unstimulated 
IFNg in pleural fluid, as measured by ELISA or 
radioimmunoassay with levels above a pre-determlned cut-off 
(varying from 1 .5-300pg/ml) In the different studies Is 89-96% 
sensitive and 96-97% specific for the diagnosis of pleural TB 

While these meta-analyses showed substantial 
heterogeneity and publication bias was possible, the observed 
accuracy of these biomarkers supports their use in the 
diagnosis of pleural TB. However, available tests for their 
measurement currently can only be performed in a laboratory. 
Therefore, we explored the feasibility of developing, In India, a 
low-cost (less than US$1), multiplexed polnt-of-care (POC) 
lateral-flow immunochromatographic test (ICT) for pleural TB 
based on IFNg and ADA. ICTs are extensively used for 
diagnosis of various Infectious diseases (e.g. syphilis, malaria) 
and have revolutionized the care of HIV by facilitating rapid 
diagnosis at the POC [18,19]. For TB no such test exists, but a 
test could conceivably improve the care of patients with pleural 
TB. Development of such a test in a high-burden country would 
also build on and expand existing capacity, allow production at 
high volume and low cost (because of lower production cost) 
and shorten delivery pathways. 


This project was a collaboration between Tulip Diagnostics 
(Goa, India), Christian Medical College ("CMC"; Vellore, India) 
and the McGill International TB Center (McGIII University, 
Montreal, Canada). The institutional review boards (IRB) of 
CMC and McGill University Health Centre (Montreal, Canada) 
approved the study. Patients who met the Inclusion criteria 
(suspected to have TB pleural effusion, at least 18 years of 
age, able and willing to give Informed consent) were explained 
the scope of the study, and written informed consent was 
obtained (IRB-approved consent form available in English, 
Hindi and Bengal). 

Derivation and validation cohort 

All patients enrolled in the study underwent thoracentesis (as 
part of their routine clinical care) in the pulmonary outpatient 
clinic or the pulmonary medical ward at CMC for evaluation of 
suspected pleural TB. When clinically indicated and safely 
feasible, a pleural biopsy was also performed. Figure 1 outlines 
the study flow and the location of testing performed at CMC 
versus Tulip. Information on demographics, co-morbldltles, 
presenting symptoms and results of routine diagnostic 
evaluation was obtained from the chart. 

For test development, discarded, frozen pleural fluid 
specimens (collected within three months prior to the study) 
from patients who had either a confirmed diagnosis of pleural 
TB or a confirmed alternative diagnosis were used (derivation 
cohort). The frozen sample was thawed and divided into two 
allquots. One was sent frozen to Tulip Diagnostics for test 
development and the remaining sample was kept at CMC 
(frozen at -80 degrees Celsius; Figure 1). 

For the validation cohort, we prospectively enrolled 107 
consecutive adult patients between August 2012 and May 2013 
at CMC. Characteristics of the validation cohort are provided In 
Table 1 . Pleural TB was suspected based on clinical symptoms 
and radiographic evidence of a pleural effusion. Clinicians used 
a LIkert scale to rate their clinical suspicion for TB (prior to test 
results becoming available) as "not likely," "likely" or "very 
likely." The collected samples were aliquoted into three 
specimens. One aliquot was tested fresh at CMC; the second 
was frozen (at -80 degrees Celsius) and stored at CMC; and 
the third was sent frozen to Tulip (Figure 1 ). 

One operator (YK) performed the IFNg ELISA/ICTs and ADA 
testing at Tulip. Two trained operators did the ICT testing at 
CMC at the POC in the clinic. For a subset of samples, both 
operators at CMC interpreted the ICT independently to assess 
inter-rater variability (Figure 1 ). The operators at the two testing 
sites (CMC and Tulip) were blinded to the reference standard 
and to the results of the tests at the other site. 

Reference standard 

All pleural fluid specimens were processed with routine 
diagnostics In the CMC microbiology laboratory. Testing 
included fluorescence smear microscopy, liquid culture 
(Mycobacterium Growth Indicator Tube, MGIT, Becton 
Dickinson, Sparks, MD, USA) and solid culture (Lowenstein 
Jensen medium) for both the derivation and validation cohort. If 
additional fluid was available, an Xpert® MTB/RIF assay was 
also done at CMC ("Xpert;" Cepheid, Sunnyvale, USA) [7]. 
Pleural tissue, if obtained, was evaluated with histopathology, 
smear microscopy, and culture. Xpert was also done on tissue 
biopsy if sufficient sample was available [6]. Results of the 
different components of the reference standard for the patients 
in the derivation and validation cohorts are provided in Table 2. 

We considered the diagnosis of pleural TB confirmed if 
smear, culture or Xpert on pleural tissue or fluid was positive 
for Mycobacterium tuberculosis (MTB), histopathology of 
pleural tissue identified granulomas, or MTB was present in 
any other respiratory sample (e.g. sputum, endobronchial or 
transbronchlal biopsy). Pleural TB was ruled out if either 
histopathology was diagnostic for malignancy or both pleural 



December 2013 | Volume 8 | Issue 12 | 885447 

A Test for IFNg in Pleural Tuberculosis 

Enrollment at CMC: 145 patients 

Reference standard: 

Fluorescence smear microscopy 

Mycobacterium Growth Indicator Tube + Lowenstein Jensen medium 
If additional fluid available: Xpert MTB/RIF (110 patients) 

If biopsy done (121 patients): histopathology, smear microscopy, culture. Xpert if extra tissue available (62 patients). 

Derivation cohort: 38 patients 

- retrospective, frozen samples 

- 12 TB, 26 alternative diagnosis 

Validation cohort: 107 patients 

- prospective, fresh samples 

- 32 TB, 65 alternative diagnosis 

10 patients excluded: 
no definite diagnosis 

2 aliquots per sample 

3 aliquots per sample 

Test development at 
Tulip Diagnostics: 

- EUSA for IFNg 

- IFNglCT 

- ADA 

Frozen sample kept 
at CMC 
- IFNglCT 

Tested at Tulip after 
1 freezing step: 

- EUSA for IFNg 

- IFNglCT 

- ADA 

Tested fresh at CMC 
- IFNglCT 
(For subset of 
samples IFNg ICT was 
interpreted by two 
operators to assess 
inter-rater variability) 

Frozen sample kept 
at CMC 

- IFNg ICT repeated 
for subset of 

Figure 1. Study flow and location of testing. Legend: CI\/!C=Christian Medical College; TB=tuberculosis; ELISA=enzyme-linked 
immunosorbent assay; ICT= lateral flow, immunochromatographic test; IFNg=interferon gamma; ADA=adenosine deaminase. 

doi: 10.1371/joumal.pone.0085447.g001 

tissue culture and histopathology were negative for TB. A 
definite alternative diagnosis (e.g. empyema or malignancy) 
was necessary for a sample to be included in the derivation 
cohort. A chart review was done for patients in the validation 
cohort with presumed false-positive results on the index test 
(but without a definite alternative diagnosis) but no additional 
TB cases were identified in follow-up. 

Lateral-flow immunochromatographic test: principle of 
test, procedure and interpretation 

The ICT utilizes the principle of immunochromatography and 
is a two-site immunoassay performed on a membrane. As the 
test sample flows through the membrane assembly of the test 
device the colored monoclonal anti-interferon gamma (IFNg) 
colloidal gold-conjugate complexes with the IFNg in the 
sample. The sample then moves further on the membrane to 
the test region, where it is immobilized by a monoclonal anti- 
IFNg antibody coated on the membrane leading to the 
formation of a pink/purple colored band, which confirms a 
positive test result. Absence of this colored band in the test 
region indicates a negative test result. The unreacted 
conjugated antibodies move further on the membrane and are 
subsequently immobilized by the goat anti-mouse antibody 
coated on the membrane at the control region, forming a pink/ 

purple band. This control band serves to validate the test 

The procedure includes four steps: 1) adding specimen (50 
pi) to the well marked "A" (see Figure 2); 2) adding sample 
running buffer to the well marked "B" (see Figure 2); 3) waiting 
30 minutes; and 4) interpreting results. 

The test is interpreted as "negative" if only one pink/purple 
band appears at the control region "C" (see Figure 2). The test 
is interpreted as "positive" if in addition to the control band a 
pink/purple band (of any intensity) appears at the test region 
"T" (see Figure 2). The test is interpreted as "invalid" if the 
control band does not appear or if the band is incomplete (i.e. 
does not span the entire width of the testing field). 

Development of the ICT - 1^' prototype 

The platform development was done at Tulip Diagnostics, 
one of the largest diagnostics manufacturers in India with 
extensive experience in the development of ICTs. Commercial 
antibodies against IFNg and ADA were obtained to first 
develop an ELISA and then a lateral flow test using standard 
principles of immunochromatography [19]. 

Of antibodies against ADA tested, no antibody was available 
that detected ADA in the quantities present in pleural fluid 
(around 30-60 lU/L measured by enzymatic activity) [3,10,12]. 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 

Table 1. Characteristics of patients in validation coliort. 




Subjects total 



Age categorized 


























Chiest pain 



Shortness of breath 



Weight loss 



Night sweats 



Unilateral effusion 



Bilateral effusion 



Pulmonary infiltrates suggestive of tuberculosis 



Cavitations 1 1 

Human immunodeficiency virus (HIV) infection 









End-stage renal disease 



History of malignancy 



Treatment with immunosuppressive medications 



Congestive heart failure 



Rheumatologic disease 



History of active tuberculosis 



Close contact with tuberculosis patient 



Clinical diagnosis at first evaluation {without test results) 

Alternative diagnosis more likely than tuberculosis 



Tuberculosis likely 



Tuberculosis very likely 



Presenting symptoms 
Radiographic findings 

*Total number of subjects in validation cohort 
doi: 10.1371/journal.pone.0085447.t001 

Thus, a decision was made early on to focus on an ICT with 
IFNg as the only biomarker. A list of ADA and IFNg antibodies 
tested, including suppliers and characteristics, is shown in 
Table S1. 

The pair of commercially available monoclonal antibodies for 
ELISA from Mabtech (Nacka Strand, Sweden) provided the 
best and most consistent results for detection of IFNg by 
ELISA. Because IFNg is commonly present in non-TB pleural 
effusions in low amounts, a cut-off for positivity was established 
at Tulip Diagnostics based on the performance of the ELISA 
(with optical density measurements) on frozen samples from in 
patients with either confirmed TB (n=12) or an alternative 
diagnosis (n=26) in the derivation cohort from CMC (Figure 1). 
A cut-off of 300pg/ml as measured by ELISA resulted in the 
best sensitivity and specificity (94% and 93%, respectively). 

Table 2. Results for different tests (within reference 
standard) performed on pleural fluid and pleural tissue. 

Diagnostic test results contributing 

to the diagnosis 


Derivation cohort 

Validation cohort 





Positive pleural tissue culture 




Positive pleural fluid culture 




MTB identified at other site 




Histopathology with granulomas 




Positive Xpert in pleural fluid 




Positive Xpert in pleural tissue 




Alternative diagnosis 




Definite malignancy by pathology 




or cytology 

Negative pleural tissue culture 
and histopathology 




Other confirmed diagnosis (e.g. 




10 patients without definite clinical diagnosis 

doi: 10.1371/joumal. pone. 0085447.1002 

Values above 300pg/ml were deemed diagnostic for pleural TB 
(of note, this value is at the higher end of the range of cut-off 
values reported in the literature) [9]. 

Subsequently, the ICT prototype (called Gammacheck, 
Figure 3) was developed at Tulip Diagnostics using antibodies 
from Hytest (Turku, Finland) as these provided the best results 
on the ICT platform (Table SI). The antibody concentrations on 
the ICT were adjusted until values above 5000pg/ml would 
result in a positive test line. A lower limit of detection was not 
feasible on the ICT platform with this antibody pair. 

Statistical analysis 

We performed all the analyses with Stata, Version 12 (Stata 
Corp, Texas, USA). The analysis and reporting followed the 
Standards for the Reporting Diagnostic Accuracy (STARD) 


Initial evaluation of 1^* prototype 

Initial results under optimized conditions at Tulip Diagnostics, 
using aliquots of the same samples that were used to define 
the cut-off for the ELISA (derivation cohort), showed a 
sensitivity of 69% (95% confidence interval (CI): 41-89) and 
specificity of 100% (95% CI: 82-100%) in comparison to the 
ELISA. Testing was then performed in the clinical setting (at 
the CMC laboratory) on frozen aliquots from the same samples 
(from the derivation cohort) as well as on newly collected fresh 
samples (from a subset of the validation cohort), following the 
standard operating procedure for the ICT provided by Tulip 
Diagnostics. While the sensitivity and specificity on fresh 
samples at CMC were lower than what was observed at Tulip 
diagnostics (sensitivity 50% [95% CI: 16-84%], specificity 90% 
[95% CI: 67-99%], the confidence interval included the values 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 

Control line 

Test line 




Figure 2. IFNg lateral flow test. (A) Initial results at Tulip with clearly defined lines indicating a positive (top) and negative 
(bottom) test. (B) Initial results at clinical evaluation site (CMC) with smearing of sample and incomplete advancement. 

doi: 10.1371/joumal.pone.0085447.g002 

observed In the original test development. However, in contrast 
to the 5% invalid results observed on initial testing at Tulip 
Diagnostics, 30% of results were invalid on testing of samples 
at CMC (Figure 2B). These invalid results were hypothesized to 
be related to the substantially higher ambient temperature 
(>30°C) and humidity at CMC (compared to the air-conditioned 
rooms at Tulip Diagnostics). In addition, we found that blood 
contamination of pleural fluid was associated with impaired flow 
when samples were tested fresh (rather than frozen). 

Optimization and evaluation of 2"" prototype 

Further optimization work was thus done at Tulip Diagnostics 
In order to develop a second prototype. An additional layer was 
integrated onto the nitrocellulose membrane to reduce the 
evaporation of chase buffer at high temperatures and 
phytohemagglutin was added to the release pad to ensure 
adequate flow even in the presence of blood. 

The evaluation of the second prototype was done at CMC on 
72 prospectively collected fresh samples from the validation 
cohort (in a non-air-conditioned room). An aliquot of each 
sample was frozen and shipped to Tulip Diagnostics for 
retesting with the prototype in addition to testing with the IFNg 
ELISA and with an enzymatic assay for ADA (ADA-MTB, Tulip 
Diagnostics) (Figure 1). 

The invalid rates of the second prototype were substantially 
improved: 3% at CMC (compared to 30% for the first prototype) 
and 0% at Tulip. In comparison with the ELISA as a reference 
standard, the performance of the ICT was as follows: sensitivity 
was 69% (95% CI: 51-83) at Tulip and 59% (95% CI: 41-75) at 
CMC and specificity was 94% (95% CI: 81-99) both at Tulip 
and at CMC. The test could not be performed on two samples 
at Tulip as the sample was too viscous after freezing and did 
not migrate along the membrane (Table 3). 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 





ML. No 55t 

LotNo. : m«>^ 

M)g. Dt 2012-04 
I Exp. Dt : 2014-03 

Manufacturod by 

Zaphyr Biomadlcals 

M • 46/4 7. Phasa III B. 
Varna Induilrlal Ealale. 
Veina. Goa - 403 722. INDIA 

Figure 3. Prototype of the IFNg lateral flow test (named Gammacheck). 

doi: 10.1371/joumal.pone.0085447.g003 

If the comparison was made to a clinical reference standard 
(definite diagnosis available for 64 patients out of 72), the 
ELISA had a sensitivity of 94% (95% CI: 79-99) and a 
specificity of 84% (95% CI: 73-92) in the validation cohort. The 
second prototype in comparison had a sensitivity of 65% (95% 
CI: 44-83) and a specificity of 89% (95% CI: 74-97) on fresh 
samples at CMC and a sensitivity of 76% (95% CI: 55-91) and 
specificity of 86% (95% CI: 71-95) after one freezing step at 
Tulip (Table 4). 

Added value of ICT for clinician decision-making 

Clinicians identified all TB patients that tested positive with 
the second prototype of the ICT as either "likely" or "very likely" 
to have TB based on clinical suspicion alone. Only for two 
patients ultimately identified as having TB did providers have a 
low suspicion of TB. These two patients also tested negative 
with the ICT. 

The second prototype of the IFNg ICT identified five 
additional cases of TB confirmed by the clinical reference 
standard that would have not been identified by standard 
biochemical analysis (of the pleural fluid) only (i.e. lymphocytic 

Table 3. Comparing 2""^ prototype of lateral flow assay to 
ELISA results. 

Location Specimen Condition 



Invalid results 

Tulip Frozen 

24/35 (69%) 

33/35 (94%) 

0/70 (0%)* 

CMC Fresh 

20/34 (59%) 

34/36 (94%) 

2/72 (3%) 

^ 2 samples with insufficient amount to be retested 
doi: 10.1371/journal. pone. 0085447.1003 

predominance and an elevated ADA). However, it also was 
falsely positive in two cases where the standard biochemical 
analysis would have been true negative. For 38 out of 64 
patients, the results of the biochemical analysis and the IFNg 
were concordant and correctly identified the diagnosis (27 true 
negative and 1 1 true positive). For 4 patients the results were 
concordant but incorrect (2 false negative and 2 false positive). 
Out of the remaining 22 patients, 15 had discordant results 
(IFNg ICT correct for 8 cases), for 2 the IFNg ICT failed and for 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 

Table 4. Comparing 2"" prototype of lateral flow assay with 
clinical diagnosis**. 

Location Specimen Condition 



Invalid results 

Tulip Frozen 

19/25 (76%) 

31/36 (86%) 

1/62 (2%)* 

CIVIC Fresh 

17/26 (65%) 

32/36 (89%) 

2/64 (3%) 

■*■ Diagnosis of TB confirmed if smear, culture or Xpert on pleural tissue or fluid was 
positive for Mycobacterium tuberculosis (MTB), histopathology of pleural tissue 
identified granulomas or MTB was present in any other sample (e.g. sputum, 
endobronchial biopsy); * 8 out of 72 patients without confirmed clinical diagnosis; * 
2 samples tested at CMC with insufficient amount to be retested at Tulip 

doi: 10.1371/journal.pone.0085447.t004 

5 patients, the biochemical analysis with cell count and ADA 
was incomplete. 

Discrepancy between results 

Fifteen percent (10 samples) of results of the second 
prototype of the ICT as performed at CIVIC versus at Tulip were 
discrepant, with the majority being positive at Tulip and 
negative at CMC. This raised concerns about the freezing step 
(that occurred prior to testing at Tulip but not for testing of fresh 
samples at CIVIC) having an effect on the amount of IFNg in the 
pleural fluid. When testing with the 2"'' prototype was repeated 
at CMC on the frozen aliquots of 32 samples (note: no surplus 
frozen sample was available for 3 out of the 10 discordant 
samples), the discordance was resolved for three out of seven 
discordant samples included in this subset of samples and no 
new discordance emerged. 

Inter-rater variability in the interpretation of the results was 
also considered. However only 1 result out of 32 read by two 
independent readers at CMC was read differently between the 
two readers (i.e. interpreting the visibility of the lines differently, 
which resulted in a positive result read by one reader and a 
negative result by the other). This one discordant result 
between the CMC operators occurred in one of the seven 
samples for which results at CMC vs. Tulip were discordant, 
suggesting a possible borderline result. Three of the seven 
discordant (CMC vs. Tulip) results remained unexplained and 
are likely due to variability in the test performance. 
Alternatively, differences in test accuracy may have been due 
to differences in the temperature at which the test was 
performed in the two sites. 

Given the limited performance of the test, and the limitations 
of further research and development possible to optimize the 
performance characteristics, a No-go decision was made and 
we decided to stop further investment into the project. 


Improved diagnostics for pleural TB are urgently needed and 
biomarkers, such as IFNg and ADA, have been identified as 
having good sensitivity and specificity. However, testing via 
currently available procedures requires a laboratory and 
technical expertise. A test that rapidly yields a result and can 

be done at the POC, even with imperfect sensitivity, would be a 
step forward [3,4]. 

In this study we aimed to develop such a test for IFNg. The 
results show that an ICT for IFNg with a lower limit of detection 
of 5 ng/ml has limited sensitivity and specificity. The high limit 
of detection could explain shortcomings in sensitivity of the ICT 
test [3,9,12]. Further research to improve the limit of detection 
of IFNg on an ICT platform should be considered. 

Prior studies that evaluated IFNg as a biomarker for pleural 
fluid also almost exclusively tested frozen samples. This could 
have resulted in cell-lysis and further release of IFNg from the 
intracellular compartment into the pleural fluid and thus 
increased levels of IFNg [17]. While we were able to increase 
the clinical sensitivity of the test with a freezing step, the 
numbers evaluated were too small and further research on 
comparing fresh and frozen samples is needed. However, this 
finding could point towards a more general problem for 
research on biomarkers that have a relatively large intracellular 
compartment that may be released through cell-lysis with a 
freeze/thaw step, as most exploratory studies are done on 
frozen samples [21]. In addition, the discrepant readings 
observed between different test sites could be related to the 
specimen itself. Pleural fluid can be a very heterogeneous 
sample and consistency can change after a freezing step, 
which potentially results in increased variability. 

Interrater variability was also considered, but not found to be 
a big concern (3%). Difficulties in reading ICT results have 
been described for HIV rapid tests, due to the subjectivity of 
test interpretation when the test yields a faint line [22,23]. 

The difference between findings in our study in comparison 
to the published literature on the performance of IFNg for the 
diagnosis of pleural TB could be reflective of both optimism 
bias and publication bias. Optimism bias for data reported in 
package inserts has been described for many TB tests [24], in 
particular those that did not undergo independent evaluation 
and approval by credible agencies such as the Federal Drug 
Administration (FDA) and World Health Organization (WHO). 
Most of the time this is observed when small numbers of 
patients are evaluated, controls are used that are not reflective 
of the population in regular clinical practice and high-risk 
groups are not included [24]. Similar reasons could also be 
implicated in the potentially optimistic results for IFNg in the 
published literature [9,11,12]. Furthermore, publication bias is 
likely contributing to the overoptimistic summary estimates in 
the meta-analyses [21,25]. 

Several studies on pleural TB diagnosis have shown that due 
to limitations of individual markers, a combination of markers 
might be more useful [26,27]. Development of an ICT for ADA 
was not feasible in our hands based on the antibodies 
available. However, alternative detection methods for ADA on a 
POC platform or alternative biomarkers could be considered in 
combination with IFNg. 

In summary, an ICT for IFNg did not achieve adequate 
sensitivity and specificity for the diagnosis of pleural TB. 
Furthermore, the test still would require a thoracentesis, which 
limits its use to hospitals and large health care centers. Further 
research should be pursued to optimize the limit of detection of 
IFNg on an ICT platform or ideally to identify biomarkers in 



December 2013 | Volume 8 | Issue 12 | e85447 

A Test for IFNg in Pleural Tuberculosis 

more easily accessible body compartments (e.g. urine or 
blood) for the diagnosis of pleural TB and EPTB in general. 

Supporting Information 

Table S1 . Antibodies tested for product deveiopment. 



We are grateful to the study coordinators (Shabana Gulam, 
Priya Samon, Amala Arumugam), and to the physicians of the 

Department of Pulmonary Medicine at the Christian Medical 
College for their support and contributions to this study. 

Author Contributions 

Conceived and designed the experiments: CMD YK SGS JSM 
AS NS TB RL NRP MP DJC. Performed the experiments: YK 
SGS DS JSM AS NS TB RL DJC. Analyzed the data: CMD YK 
SGS JSM AS NS TB RL NRP MP DJC. Contributed reagents/ 
materials/analysis tools: CMD YK SGS JSM AS NS MP DJC. 
Wrote the manuscript: CMD YK SGS JSM DS AS NS TB RL 


1. Selbert AF, Haynes J Jr., Middleton R, Bass JB Jr. (1991) Tuberculous 
pleural effusion. Twenty-year experience. Chest 99: 883-886. doi: 
10.1378/chest.99.4.883. PubMed: 1901261. 

2. Udwadia ZF, Sen T (2010) Pleural tuberculosis: an update. Cun- Opin 
Pulm Med 16: 399-406. doi:10.1097/MCP.0b013e328339cf6e. 
PubMed: 20531085. 

3. Dinnes J, Deeks J, Kunst H, Gibson A, Cummins E et al. (2007) A 
systematic review of rapid diagnostic tests for the detection of 
tuberculosis Infection. Health Technol Assess 11: 1-314. PubMed: 

4. Porcel JM (2009) Tuberculous pleural effusion. Lung 187: 263-270. dol: 
10.1007/S00408-009-9165-3. PubMed: 19672657. 

5. Pal M, Flores LL, Hubbard A, Riley LW, Colford JM Jr. (2004) Nucleic 
acid amplification tests in the diagnosis of tuberculous pleuritis: a 
systematic review and meta-analysis. BMC Infect Dis 4: 6. doi: 
10.1186/1471-2334-4-6. PubMed: 15102325. 

6. Christopher DJ, Schumacher SG, Michael JS, Luo R, Balamugesh T et 
al. (2013) Performance of Xpert MTB/RIF on pleural tissue for the 
diagnosis of pleural tuberculosis. Eur Respir J 42: 1427-1429. doi: 
10.1183/09031936.00103213. PubMed: 23900990. 

7. Friedrich SO, von Groote-Bidlingmaier F, Diacon AH (2011) Xpert 
MTB/RIF assay for diagnosis of pleural tuberculosis. J Clin Microbiol 
49: 4341-4342. doi:10.1128/JCM.05454-11. PubMed: 21998430. 

8. Tortoli E, Russo C, Piersimoni C, Mazzola E, Dal Monte P et al. (2012) 
Clinical validation of Xpert MTB/RIF for the diagnosis of extrapulmonary 
tuberculosis. Eur Respir J 40: 442-447. doi: 
10.1183/09031936.00176311. PubMed: 22241741. 

9. Jiang J, Shi HZ, Liang QL, Qin SM, Qin XJ (2007) Diagnostic value of 
interferon-gamma in tuberculous pleurisy: a metaanalysis. Chest 131: 
1133-1141. doi:10.1378/chest06-2273. PubMed: 17426220. 

10. Liang QL, Shi HZ, Wang K, Qin SM, Qin XJ (2008) Diagnostic accuracy 
of adenosine deaminase in tuberculous pleurisy: a meta-analysis. 
Respir Med 102: 744-754. doi:10.1016/j.rmed.2007.12.007. PubMed: 

11. Goto M, Noguchi Y, Koyama H, Hira K, Shimbo T et al. (2003) 
Diagnostic value of adenosine deaminase in tuberculous pleural 
effusion: a meta-analysis. Ann Clin Biochem 40: 374-381. doi: 
10.1258/000456303766477011. PubMed: 12880538. 

12. Greco S, Girardi E, Masciangelo R, Capoccetta GB, SaltinI C (2003) 
Adenosine deaminase and Interferon gamma measurements for the 
diagnosis of tuberculous pleurisy: a meta-analysis. Int J Tuberc Lung 
Dis 7: 777-786. PubMed: 12921155. 

13. Perez-Rodriguez E, Perez Walton IJ, Sanchez Hernandez JJ, Panares 
E, Rubi J et al. (1999) ADAI/ADAp ratio in pleural tuberculosis: an 
excellent diagnostic parameter in pleural fluid. Respir Med 93: 816-821. 
doi:10.1016/S0954-6111(99)90267-6. PubMed: 10603631. 

14. Buel E, MacQuarrie R (1981) Purification of adenosine deaminase from 
human red blood cells. Prep Biochem 11: 363-380. doi: 
10.1080/00327488108065529. PubMed: 7312831. 

15. Ungerer JP, Oosthuizen HM, Bissbort SH, Vermaal< WJ (1992) Serum 
adenosine deaminase: isoenzymes and diagnostic application. Clin 
Chem 38: 1322-1326. PubMed: 1623598. 

16. Ruddy M, Balabanova Y, Graham C, Fedorin I, Malomanova N et al. 
(2005) Rates of drug resistance and risk factor analysis in civilian and 
prison patients with tuberculosis in Samara Region, Russia. Thorax 60: 
130-135. doi:10.1136/thx.2004.026922. PubMed: 15681501. 

17. Schroder K, Hertzog PJ, Ravasi T, Hume DA (2004) Interferon-gamma: 
an overview of signals, mechanisms and functions. J Leukoc Biol 75: 
163-189. PubMed: 14525967. 

18. Pant Pal N, Balram B, Shivkumar S, Martinez-Cajas JL, Claessens C et 
al. (2012) Head-to-head comparison of accuracy of a rapid point-of- 
care HIV test with oral versus whole-blood specimens: a systematic 
review and meta-analysis. Lancet Infect Dis 12: 373-380. doi:10.1016/ 
S1473-3099(1 1)70368-1. PubMed: 22277215. 

19. Posthuma-Trumpie GA, Korf J, van Amerongen A (2009) Lateral flow 
(immuno)assay: its strengths, weaknesses, opportunities and threats. A 
literature survey. Anal Bioanal Chem 393: 569-582. dol:10.1007/ 
s002 16-008-2287-2. PubMed: 18696055. 

20. Bossuyt PM, Reltsma JB, Bruns DE, Gatsonis OA, Glasziou PP et al. 
(2003) The STARD statement for reporting studies of diagnostic 
accuracy: explanation and elaboration. Ann Intern Med 138: W1-12. 
doi: 1 0.7326/0003-481 9-1 38-1 -200301 070-0001 2-w1 . PubMed: 

21. loannidis JP, Panagiotou OA (2011) Comparison of effect sizes 
associated with biomarkers reported in highly cited individual articles 
and In subsequent meta-analyses. JAMA 305: 2200-2210. dol:10.1001/ 
jama.201 1.713. PubMed: 21632484. 

22. Klarkowski DB, Wazome JM, Lokuge KM, Shanks L, Mills OF et al. 
(2009) The evaluation of a rapid in situ HIV confirmation test In a 
programme with a high failure rate of the WHO HIV two-test diagnostic 
algorithm. PLOS ONE 4: e4351. doi:10.1371/Journal.pone.0004351. 
PubMed: 19197370. 

23. Lee VJ, Tan SC, Earnest A, Seong PS, Tan HH et al. (2007) User 
acceptability and feasibility of self-testing with HIV rapid tests. J Acquir 
Immune Defic Syndr45: 449-453. doi:10.1097/QAI.0b013e318095a3f3. 
PubMed: 17554213. 

24. Denkinger CM, Grenler J, Minion J, Pai M (2012) Promise versus 
reality: optimism bias In package inserts for tuberculosis diagnostics. J 
Clin Microbiol 50: 2455-2461. doi:10.1 128/JCM.00842-12. PubMed: 

25. loannidis JP (2008) Why most discovered true associations are 
inflated. Epidemiology 19: 640-648. doi:10.1097/EDE. 
0b013e31818131e7. PubMed: 18633328. 

26. Porcel JM, Vivos M (2003) Differentiating tuberculous from malignant 
pleural effusions: a scoring model. Med Sol Monit 9: CR175-180 

27. Villegas MV, Labrada LA, Saravia NG (2000) Evaluation of polymerase 
chain reaction, adenosine deaminase, and Interferon-gamma In pleural 
fluid for the differential diagnosis of pleural tuberculosis. Chest 118: 
1355-1364. doi:10.1378/chest.118.5.1355. PubMed: 11083686. 



December 2013 | Volume 8 | Issue 12 | 885447