Articles

Clinical utility of existing and second-generation

interferon-γ release assays for diagnostic evaluation of
tuberculosis: an observational cohort study
Hilary S Whitworth, Amarjit Badhan*, Aime A Boakye*, Yemisi Takwoingi*, Melanie Rees-Roberts*, Christopher Partlett, Heather Lambie,
John Innes, Graham Cooke, Marc Lipman, Christopher Conlon, Derek Macallan, Felix Chua, Frank A Post, Martin Wiselka, Gerrit Woltmann,
Jonathan J Deeks, Onn Min Kon†, Ajit Lalvani†, on behalf of the Interferon-γ Release Assays for Diagnostic Evaluation of Active Tuberculosis
study group‡

Summary
Background The clinical utility of interferon-γ release assays (IGRAs) for diagnosis of active tuberculosis is unclear, Lancet Infect Dis 2019;
although they are commonly used in countries with a low incidence of tuberculosis. We aimed to resolve this clinical 19: 193–202

uncertainty by determining the accuracy and utility of commercially available and second-generation IGRAs in the Published Online
January 14, 2019
diagnostic assessment of suspected tuberculosis in a low-incidence setting.
http://dx.doi.org/10.1016/
S1473-3099(18)30613-3
Methods We did a prospective cohort study of adults with suspected tuberculosis in routine secondary care in England. This online publication has
Patients were tested for Mycobacterium tuberculosis infection at baseline with commercially available (T-SPOT.TB and been corrected; the first
QuantiFERON-TB Gold In-Tube [QFT-GIT]) and second-generation (incorporating novel M tuberculosis antigens) correction appeared at
thelancet.com/infection on
IGRAs and followed up for 6–12 months to establish definitive diagnoses. Sensitivity, specificity, positive and negative
January 30, 2019. The second
likelihood ratios, and predictive values of the tests were determined. correction appeared on
January 31, 2019
Findings Of the 1060 adults enrolled in the study, 845 were included in the analyses and 363 were diagnosed with See Comment page 121
tuberculosis. Sensitivity of T-SPOT.TB for all tuberculosis diagnosis, including culture-confirmed and highly probable *Authors contributed equally
cases, was 81·4% (95% CI 76·6–85·3), which was higher than QFT-GIT (67·3% [62·0–72·1]). Second-generation IGRAs †Authors contributed equally
had a sensitivity of 94·0% (90·0–96·4) for culture-confirmed tuberculosis and 89·2% (85·2–92·2) when including ‡Members of study group listed
highly probable tuberculosis, giving a negative likelihood ratio for all tuberculosis cases of 0·13 (95% CI 0·10–0·19). at the end of the Article
Specificity ranged from 86·2% (95% CI 82·3–89·4) for T-SPOT.TB to 80·0% (75·6–83·8) for second-generation IGRAs. Tuberculosis Research Centre,
National Heart and Lung
Interpretation Commercially available IGRAs do not have sufficient accuracy for diagnostic evaluation of suspected Institute (H S Whitworth PhD,
A Badhan MSc, A A Boakye MSc,
tuberculosis. Second-generation tests, however, might have sufficiently high sensitivity, low negative likelihood ratio, M Rees-Roberts PhD,
and correspondingly high negative predictive value in low-incidence settings to facilitate prompt rule-out of H Lambie MSc, Prof O M Kon MD,
tuberculosis. Prof A Lalvani DM) and National
Institute for Health Research
Health Protection Research
Funding National Institute for Health Research. Unit in Respiratory Infections
(A Badhan, A A Boakye,
Copyright © 2019. Elsevier Ltd. All rights reserved. M Rees-Roberts, Prof O M Kon,
Prof A Lalvani), Imperial College
London, London, UK;
Introduction although these tests have high specificity (which is Department of Clinical
Prompt diagnosis and treatment of active tuberculosis important to rule in tuberculosis), they have insufficient Research, London School of
are essential for optimal patient outcomes and sensitivity to rule out tuberculosis and require clinical Hygiene and Tropical Medicine,
London, UK (H S Whitworth);
preventing onward transmission in the community and specimens from anatomical disease sites, which are
Institute of Applied Health
in health-care facilities.1 However, diagnostic assessment often obtained with resource-intensive invasive Research, University of
of suspected tuberculosis can be lengthy, costly, and procedures.4 A blood test of high diagnostic sensitivity Birmingham, Birmingham, UK
burdensome for patients and health-care systems,2 could help to promptly (eg, in 24 h) triage patients at (Y Takwoingi PhD, C Partlett PhD,
Prof J J Deeks PhD); Centre for
often resulting in substantial delays in diagnosis and clinical presentation (appendix), which would address a
Health Services Studies,
treatment of other diseases in cases when suspected major unmet clinical need and has been prioritised by University of Kent, Canterbury,
tuberculosis is eventually ruled out. Therefore, WHO.5 Given the paucibacillary nature of most cases of UK (M Rees-Roberts); Heart of
improving and accelerating diagnostic evaluation culture-negative tuberculosis, such a test would England National Health
Service (NHS) Foundation
remains a clinical and public health priority in high- probably be based on measurement of immune Trust, Birmingham, UK
income, low-incidence countries, as well as in high- responses to Mycobacterium tuberculosis rather than (J Innes FRCP); Department of
burden regions. Over the past decade, advances in direct detection of the bacteria or nucleic acids. Infectious Diseases, St Mary’s
molecular diagnostics, such as the GeneXpert assay Interferon-γ release assays (IGRAs) are regulatory- Hospital, Imperial College
Healthcare Trust, London, UK
(Cepheid, Sunnyvale, CA, USA), have improved the approved, immune-based blood tests for detecting (Prof G Cooke PhD); Department
speed and accuracy of microbiological diagnosis and M tuberculosis infection. By measuring T-cell responses of Respiratory Medicine, Royal
enabled prediction of antibiotic susceptibility.3 However, to two strongly immunogenic but highly specific Free London NHS Foundation

www.thelancet.com/infection Vol 19 February 2019 193

 Articles

Trust, London, UK
(M Lipman MD); University Research in context
College London Respiratory,
Division of Medicine, Evidence before this study tuberculosis-incidence setting. Because the study was done at
University College London, Although the utility of interferon-γ release assays (IGRAs) for the multiple centres in England, and patients representing the full
London, UK (M Lipman); diagnosis of active tuberculosis is unclear, their use in clinical natural clinical spectrum of tuberculosis were recruited, the
Nuffield Department of
Medicine, Oxford University
practice is common. A comprehensive systematic review and results are generalisable to other high-income, low-incidence
Hospitals NHS Trust, Oxford, meta-analysis published in 2011 described data from studies settings. By showing that existing IGRAs have no useful role in
UK (Prof C Conlon MD); published between January, 2001, and November, 2009, that the diagnosis of active tuberculosis, this study aims to resolve
Infection Care Group evaluated the diagnostic accuracy of IGRAs for active tuberculosis. a major clinical uncertainty and represents a substantial new
(Prof D Macallan PhD) and
Department of Respiratory
Therefore, we searched PubMed for original research studies high-quality component of the evidence base. Simultaneous
Medicine (F Chua PhD), published in any language between Dec 1, 2009, and June 30, evaluation of second-generation IGRAs (incorporating novel
St George’s University 2018, using the search terms (Tuberculosis OR TB) AND (active Mycobacterium tuberculosis antigens) identifies a potentially
Hospitals NHS Foundation
OR disease) AND (interferon-γ release assay OR T-SPOT.TB OR useful high-sensitivity triage test that meets, we believe, a
Trust, London, UK; Institute of
Infection and Immunity, QuantiFERON) AND (diagnosis OR evaluation OR rule-in OR major unmet clinical need.
St George’s, University of rule-out). The evidence base suggests that commercially available
Implications of all the available evidence
London, London, UK IGRAs have insufficient specificity to rule in tuberculosis and
(Prof D Macallan); Department Existing IGRAs do not have sufficient sensitivity or negative
insufficient sensitivity to rule out tuberculosis. However, this
of Sexual Health and HIV, predictive value to rule out a diagnosis of tuberculosis.
King’s College Hospital NHS finding is derived primarily from studies that are either small, low
They also have low specificity and, as a result, are unable to
Foundation Trust, London, UK quality, or not representative of patient populations in routine
rule in a diagnosis of tuberculosis. Thus, existing IGRAs do not
(Prof F A Post PhD); Department clinical practice. Only one large prospective cohort study
of Infection and Tropical have a clinically useful role in the diagnostic investigation of
embedded in routine practice was identified, but in a
Medicine, University Hospitals tuberculosis. The finding that second-generation IGRAs might
of Leicester NHS Trust, high-incidence setting of tuberculosis. Thus, 15 years after the
have sufficiently high sensitivity, low negative likelihood ratio,
Leicester, UK introduction of IGRAs, the ability of policy makers in
and high negative predictive value to be used as a triage test
(Prof M Wiselka PhD); and low-incidence settings to generate recommendations and
Department of Infection, to help rule out a diagnosis of tuberculosis within 24 h (the
guidelines for the use of IGRAs to diagnose active tuberculosis is
Immunity and Inflammation, time needed to obtain the test result) indicates a clinically
Respiratory Biomedical still hampered by a paucity of reliable and informative evidence.
useful role for this novel test and provides the basis for
Research Centre, Institute for
Lung Health, University of Added value of this study evidence-based guidelines on its use in low-incidence settings
Leicester, Leicester, UK This study is the largest prospective study to investigate the once it is widely available postlicensure.
(G Woltmann MD) utility of IGRAs for the diagnosis of active tuberculosis in a low
Correspondence to:
Prof Ajit Lalvani, Tuberculosis
Research Centre, National Heart
and Lung Institute, Imperial M tuberculosis antigens (ESAT-6 and CFP-10), they are not of T-SPOT.TB and QuantiFERON-TB Gold In-Tube
College London, London confounded by previous BCG vaccination and provide (QFT-GIT) tests was published in 2018, the study was
W2 1PG, UK
higher diagnostic specificity than the tuberculin skin done in a setting with a high incidence of tuberculosis.12
[email protected]
test.6 Given that M tuberculosis infection is a prerequisite Prospective cohort studies done in low-incidence settings
for tuberculosis disease, a negative IGRA result could have been substantially smaller.1,8
See Online for appendix
potentially rule out a diagnosis of tuberculosis (ie, exclude Given the shortfalls associated with available
tuberculosis from the differential diagnosis), although tuberculosis diagnostics, IGRAs continue to be used
previous evidence suggests the sensitivity of available widely in clinical practice in the UK, albeit resulting in
IGRAs might be insufficient to fulfil this triage complexities and challenges in interpretation of the
function.1,7–9 results.11 Therefore, a large-scale prospective head-to-
Although established as the new standard of care for head comparison of diagnostic performance of IGRAs in
diagnosing latent tuberculosis infection, IGRAs are not routine practice is required to conclusively define which,
recommended for the diagnosis of active tuberculosis if any, clinical role they have in the diagnosis of active
other than in specific scenarios (eg, in children), with tuberculosis, allowing development of evidence-based
caveats around interpretation of the results and the and authoritative recommendations in this setting.
expertise required.10,11 However, development of definitive Discovery of other highly specific M tuberculosis
recommendations has been hindered by a paucity of antigens that are as strongly immunogenic as ESAT-6
robust and informative evidence. Most studies of the and CFP-10 presents an opportunity to develop second-
diagnostic accuracy of IGRAs in active tuberculosis, to our generation IGRAs of higher sensitivity.13,14 Furthermore,
knowledge, are retrospective reviews of hospital records they might allow development of an ESAT-6-free IGRA
and tuberculosis registry data or small case-control for application in populations vaccinated with new
studies, typically not representative of the heterogeneous ESAT-6-based tuberculosis vaccines, as previously
patient population seen in clinical practice. Although one described.15 Evidence from previous studies suggests that
large (n=746) prospective cohort study embedded in the adaptation of existing IGRAs with these novel
routine practice and including a head-to-head comparison antigens is possible,1,14,16 but no large-scale prospective

194 www.thelancet.com/infection Vol 19 February 2019

 Articles

clinical evaluation of this novel approach has been done services, before a final diagnosis was made, and a wide
in routine practice in a low tuberculosis incidence spectrum of pretest probabilities for active tuberculosis
setting. were included. Exclusion criteria were restricted to age
Therefore, we sought to evaluate the clinical utility younger than 16 years and inability or unwillingness to
of existing IGRAs—T-SPOT.TB (Oxford Immunotec, provide informed consent. Centres were selected to ensure
Abingdon, UK) and QFT-GIT (Qiagen, Hilden, Ger-​ the population was representative of various ethnic groups
many)—and second-generation IGRAs incorporating and comorbidities.
novel M tuberculosis antigens in patients presenting with
suspected tuberculosis in English hospitals. Participant enrolment and follow-up
Participants were first observed by research nurses at
Methods enrolment. Following consent, a baseline blood sample
Study design was drawn and data were collected in a case report form on
We did a prospective, multicentre, cohort study in routine the demographics and medical history of the participant
clinical practice in England. A within-patient design was and on investigations performed during routine diagnostic
used to compare test accuracy: blood samples from each work-up. Participants were followed up for 2 months and
study participant were tested with all IGRAs, with the 6 months thereafter, with data collected on any subsequent
presence or absence of active tuberculosis verified with a investigations, test results, diagnoses, and response to
composite reference standard (table 1).1 This design tuberculosis treatment if initiated. Patients with a definitive
minimises between-patient variability. The study was non-tuberculosis diagnosis who were discharged from
approved by Camden and Islington National Research routine care were not required to attend follow-up visits,
Ethics Committee (reference 11/H0722/8). All participants but when necessary, data were collected from hospital
provided written, informed consent. The study protocol is records up to 12 months after enrolment to identify final For the study protocol see
available online and a standards for reporting studies of diagnoses made by hospital clinicians. https://njl-admin.nihr.ac.uk/
document/download/2006627
diagnostic accuracy checklist is provided in the appendix.
Diagnosis and diagnostic categorisation
Study participants Participants were investigated in routine practice under
Adult inpatients and outpatients presenting with suspected the direction of the infectious disease or respiratory
active tuberculosis (based on signs and symptoms assessed medicine attending physician. After completion of
by the attending hospital clinician) were consecutively follow-up in this routine hospital setting, participants’
enrolled from ten National Health Service hospitals in five final diagnoses were verified with a composite reference
English cities (London, Slough, Oxford, Leicester, and standard1 by a panel of four or more respiratory medicine
Birmingham). Patients were enrolled at presentation to and infectious disease clinicians specialising in
infectious disease and respiratory medicine secondary care tuberculosis. The panel assessed anonymised clinical

Criteria Number of patients

(n/N)
1: Culture-confirmed tuberculosis* Microbiological culture of Mycobacterium tuberculosis, and clinical and radiological findings 261/845 (31%)
suggestive of tuberculosis
2: Highly probable tuberculosis* Clinical and radiological features highly suggestive of tuberculosis and unlikely to be caused 102/845 (12%)
by other diseases, a decision to treat made by a clinician, appropriate response to therapy,
and histological evidence if available
3: Clinically indeterminate diagnosis Final diagnosis of tuberculosis neither highly probable nor reliably excluded 43/845 (5%)
4: Active tuberculosis excluded
4A: Inactive tuberculosis Stable chest x-ray changes, TST positive† (if available), bacteriologically negative 7/845 (1%)
(if available), and no clinical evidence of active disease
4B: One or more risk factors for TST positive†, bacteriologically negative (if available), and no clinical evidence of active 48/845 (10%)
tuberculosis exposure‡, TST positive† disease
4C: One or more risk factors for History of exposure to tuberculosis and TST negative (if available) 267/845 (32%)
tuberculosis exposure‡, TST negative
4D: No risk factors for tuberculosis No history of exposure to tuberculosis and TST negative (if available) 117/845 (14%)
exposure‡, TST negative

TST=tuberculosis skin test. *M tuberculosis culture is the gold standard for diagnosis of active tuberculosis; however, given that culture does not detect all tuberculosis
cases, our previously validated standard reference includes a second category for culture-negative but highly probable active tuberculosis diagnoses, made on the basis of
other available evidence.1 †TST by use of Mantoux test, with a threshold ≥15 mm considered positive. ‡Risk factors for exposure to tuberculosis: recent exposure to a
patient with active tuberculosis, born in country of high prevalence, or belonging to an ethnic group with a very high prevalence of tuberculosis (incidence >100 per
100 000 people).

Table 1: Predefined criteria for case definitions and diagnostic categories1

www.thelancet.com/infection Vol 19 February 2019 195

 Articles

1060 patients with suspected tuberculosis

recruited from ten sites

215 excluded
99 previous tuberculosis diagnosis*
66 excluded by investigators†
39 lost to follow-up
8 withdrew consent
3 deaths

845 included in the analyses

261 culture-confirmed tuberculosis 102 highly probable tuberculosis 43 clinically intermediate cases 439 active tuberculosis cases
cases cases 41 TSPOT.TB and QFT-GIT excluded
245 TSPOT.TB and QFT-GIT 99 TSPOT.TB and QFT-GIT 0 TSPOT.TB only 420 TSPOT.TB and QFT-GIT
1 TSPOT.TB only 0 TSPOT.TB only 1 QFT-GIT only 3 TSPOT.TB only
7 QFT-GIT only 2 QFT-GIT only 1 IGRA result not available 5 QFT-GIT only
8 IGRA results not available 1 IGRA result not available 11 IGRA results not available

Figure: Study selection for patients with suspected active tuberculosis

QFT-GIT=QuantiFERON-TB Gold In-Tube. IGRAs=interferon-γ release assays. *Patients with previously diagnosed tuberculosis were excluded from analyses because results
from IGRAs cannot be reliably interpreted in these patients; this decision was made by the expert diagnostic panel and study management group in consultation with the
independent study steering committee before unblinding of IGRA and second-generation IGRA results. †Upon advice from the study steering committee, and following
consultation between the study management group and data management groups, a total of 66 patients were excluded from analyses; patients recruited from one study
site (n=40) were excluded because they all had diagnoses of confirmed or highly probable tuberculosis (categories 1 and 2) due to an error of implementation of
recruitment criteria at this site (ie, patients with suspected tuberculosis were not being recruited). A further subset of patients (n=26) were, upon review, considered by the
expert diagnostic panel to be ineligible (before unblinding IGRA results) on the basis that they were being investigated for tuberculosis (incidental atypical chest x-ray,
known contact with active tuberculosis, or screening for anti-tumour necrosis factor treatment), but did not present with symptoms or signs suggestive of tuberculosis.

data (patient demographics, medical history, tuberculosis appendix. The laboratory scientists were masked to all
symptoms, previous tuberculosis information, history of clinical information, diagnoses, and personal identifiers.
exposure to tuberculosis, current medication, HIV
status, relevant clinical correspondence, test results Statistical analysis
during diagnosis and follow-up, and any other relevant This study was powered to detect a 10% difference in
clinical information) while being masked to all IGRA sensitivity between T-SPOT.TB and QFT-GIT, assuming a
results (including IGRAs done as part of routine practice sensitivity of 85% for T-SPOT.TB and of 75% for
at the recruitment sites). QFT-GIT.1,7–9 Accounting for the paired nature of the data
Diagnoses of all participants were categorised into the and assuming independence of errors,17 855 patients (after
following groups (table 1): definite tuberculosis (category 1), loss to follow-up, withdrawal, or exclusion because of
highly probable tuberculosis (category 2), clinically in- missing or invalid index or reference test results) were
determinate diagnosis (category 3), and non-tuberculosis required to detect this difference at the 5% significance
(category 4). Category 4 participants were subdivided on level (two-tailed) with 90% power, based on a predicted
the basis of risk factors for latent tuberculosis infection. 40% prevalence of active tuberculosis in the study
Final diagnoses and diagnostic categories were determined population. We calculated sensitivity, specificity, positive
by consensus across the panel. and negative predictive values, and positive and negative
likelihood ratios for each test. 95% CIs were calculated
Laboratory procedures with the Wilson method for proportions,18,19 and likelihood
Blood samples (35 mL) were collected from all participants ratios with the method by Simel and colleagues.20 All
at enrolment in heparinised and QFT-GIT blood collection patients in diagnostic categories 1, 2, and 4 were included
tubes. QFT-GIT and T-SPOT.TB were done and interpreted in analyses (table 1); data for patients in category 3 were
in real time at the Tuberculosis Research Centre reported but not included in the analyses.
(Imperial College London, London, UK) according to Patients with indeterminate IGRA or borderline
the manufacturer’s instructions and as described by T-SPOT.TB results were excluded from primary analyses
Whitworth and colleagues.6 The second-generation IGRA but included as test positives in sensitivity analyses.
used the T-SPOT.TB platform and incorporated ESAT-6, Sensitivity analyses were also done to investigate the
CFP-10, and Rv3615c; the ESAT-6-free IGRA incorporated effects of excluding category 2 patients on IGRA sensitivity
CFP-10, Rv3615c, and Rv3879c. Further details on assay and of excluding category 4A–C patients on IGRA
methods and interpretation of results are provided in the specificity. To compare the accuracy of two IGRAs, we fit

196 www.thelancet.com/infection Vol 19 February 2019

 Articles

Culture-confirmed Highly probable Clinically Active tuberculosis Total

tuberculosis (n=261) tuberculosis (n=102) indeterminate (n=43) excluded (n=439) (n=845)
Clinical setting
Outpatient 171 (66%) 72 (71%) 32 (74%) 269 (61%) 544 (64%)
Inpatient 90 (34%) 30 (29%) 11 (2·6%) 170 (3·9%) 301 (3·6%)
Age, years 32 (26–42) 36 (29–45) 38 (27–56) 44 (33–56) 38 (30–51)
Male 177 (68%) 53 (52%) 21 (4·9%) 250 (57%) 501 (59%)
Female 84 (32%) 49 (48%) 22 (51%) 189 (43%) 344 (41%)
Ethnic origin
Indian subcontinent 167 (64%) 61 (60%) 16 (37%) 168 (38%) 412 (49%)
Black 50 (19%) 22 (22%) 10 (23%) 102 (23%) 184 (22%)
White 22 (8%) 9 (9%) 12 (2·8%) 126 (29%) 169 (20%)
Asian 16 (6%) 6 (6%) 5 (12%) 14 (3%) 41 (5%)
Middle Eastern 4 (2%) 0 0 12 (3%) 16 (2%)
Mixed 1 (<1%) 4 (3·9%) 0 8 (2%) 13 (2%)
Hispanic 1 (<1%) 0 0 7 (2%) 8 (1%)
Unknown 0 0 0 2 (<1%) 2 (<1%)
Years in the UK 4·8 (2·3–10·7) 6·1 (3·1–16·3) 10·6 (5·8–34·5) 13·1 (6·4–32·1) 8·3 (3·3–20·4)
Profession*
Paid employment 130 (50%) 52 (51%) 21 (4·9%) 214 (4·9%) 417 (49%)
Unemployed 62 (24%) 24 (24%) 16 (37%) 164 (37%) 266 (31%)
Student 50 (19%) 13 (13%) 3 (7%) 26 (6%) 92 (11%)
Health-care or laboratory worker 16 (6%) 9 (9%) 2 (5%) 24 (5%) 51 (6%)
Social or prison worker 1 (<1%) 1 (1%) 0 2 (<1%) 4 (<1%)
Sex worker 0 1 (1%) 0 2 (<1%) 3 (<1%)
Unknown 2 (1%) 2 (2%) 1 (2%) 7 (2%) 12 (1%)
Height, m 1·7 (1·6–1·8) 1·7 (1·6–1·8) 1·6 (1·6–1·7) 1·7 (1·6–1·8) 1·7 (1·6–1·8)
Weight, kg 63 (55–72) 64 (54–75) 71 (57–85) 68 (58–79) 65 (57–77)
Body-mass index, kg/m² 22 (20–25) 22 (21–26) 24 (20–32) 24 (21–28) 23 (20–27)
BCG vaccinated 194 (74%) 79 (77%) 36 (8·4%) 340 (77%) 649 (77%)
BCG scar visible
Yes 172 (6·6%) 72 (7·1%) 29 (67%) 283 (64%) 556 (66%)
No 12 (5%) 3 (3%) 3 (7%) 19 (4%) 37 (4%)
Unknown 16 (6%) 8 (8%) 6 (14%) 44 (10%) 74 (9%)
Data not available 61 (23%) 19 (1·9%) 5 (1·2%) 93 (21%) 178 (21%)
Known contact with tuberculosis 70 (27%) 25 (25%) 12 (28%) 83 (19%) 190 (22%)
Other pre-existing conditions or comorbidities†
None 169 (65%) 61 (60%) 19 (44%) 169 (38%) 418 (49%)
HIV infected 13 (5%) 12 (12%) 2 (4%) 108 (25%) 135 (16%)
Diabetes 22 (8%) 5 (5%) 8 (19%) 53 (12%) 88 (10%)
Asthma 12 (5%) 5 (5%) 4 (9%) 50 (11%) 71 (8%)
Cancer 1 (<1%) 1 (1%) 0 12 (3%) 14 (2%)
Chronic or end-stage kidney disease 5 (2%) 1 (1%) 2 (5%) 4 (1%) 12 (1%)
Hepatitis C 1 (<1%) 1 (1%) 0 10 (2%) 12 (1%)
Hepatitis B 5 (2%) 1 (1%) 0 5 (1%) 11 (1%)
Organ transplant 0 0 0 2 (<1%) 2 (<1%)
Sarcoidosis 1 (<1%) 0 0 0 1 (<1%)
Other 74 (28%) 37 (36%) 20 (47%) 228 (52%) 359 (42%)
Data are n (%) or median (IQR). *Some patients had more than one profession. †Some patients had multiple comorbidities.

Table 2: Demographics and clinical characteristics of participants

separate generalised estimating equation models for nature of the data while allowing use of all available data if
patients with active tuberculosis to estimate differences in test results were missing for either IGRA. We computed
sensitivity and without active tuberculosis to investigate ratios of sensitivities (relative sensitivity) and specificities
differences in specificity. This approach exploits the paired (relative specificity) from the generalised estimating

www.thelancet.com/infection Vol 19 February 2019 197

 Articles

equation models using a postestimation procedure, with the cohort was 38 years (IQR 30–51); over half of the
95% CIs calculated with the delta method. Analyses were patients were men, and almost half were from the Indian
done with Stata (version 13.0). subcontinent. One or more comorbidities were reported
in 427 (51%) of 845 participants (table 2). Medications at
Role of the funding source presentation are shown in the appendix. The most
The funder of the study had no role in study design, data common symptoms reported at presentation were cough,
collection, data analysis, data interpretation, or writing of weight loss, and lethargy (appendix).
the report. The corresponding author had full access to 363 (43%) of 845 patients had a final diagnosis of active
all the data in the study and had final responsibility for tuberculosis (categories 1 and 2; table 1). Of these,
the decision to submit for publication. 129 (36%) had pulmonary tuberculosis, 189 (52%) had
extrapulmonary tuberculosis, and 45 (12%) had both
Results (table 3); most patients had lymph node involvement. Of
Between Nov 25, 2011, and Aug 31, 2013, 1060 patients with 261 patients who had drug-susceptibility testing done on
suspected active tuberculosis provided consent and were M tuberculosis isolates, 21 (6%) had drug-resistant
enrolled. Patients with a history of tuberculosis diagnosis tuberculosis and one (<1%) had multidrug-resistant
(n=99) were excluded from analyses, as in previous tuberculosis. Tuberculosis was excluded (category 4) in
studies.12 Additionally, 116 patients were also excluded 439 (52%) of 845 patients (table 1). These patients were
(figure), giving a final study population of 845 patients. subclassified according to risk factors for latent
Demographic and clinical characteristics for the final tuberculosis infection or inactive tuberculosis into
study population are shown in table 2. The median age of categories 4 A–D in decreasing likelihood of having
M tuberculosis infection (table 1).1 The most common
Confirmed or highly probable Active tuberculosis non-tuberculosis diagnoses are listed in table 3. Only
tuberculosis (n=363) excluded (n=439)* 43 (5%) of 845 patients were classified as clinically
All cases 363 (100%) ·· indeterminate (category 3).
Pulmonary 129 (36%) ·· TSPOT.TB results were available for 809 (96%) of
Extrapulmonary 189 (52%) ·· 845 study participants and QFT-GIT results for 820 (97%);
Pulmonary and extrapulmonary 45 (12%) ·· data from both tests were available for 805 (95%) patients
Site of disease†
(figure). Diagnostic sensitivity, specificity, positive and
Lungs 174 (48%) ··
negative predictive values, and positive and negative
Lymph node 154 (42%) ··
likelihood ratios are shown in table 4, with a cross-
tabulation of T-SPOT.TB and QFT-GIT results in patients
Pleura 26 (7%) ··
with active tuberculosis and non-tuberculosis diagnoses
Spine 16 (4%) ··
provided in the appendix. Sensitivity of T-SPOT.TB was
Miliary tuberculosis (disseminated) 11 (3%) ··
84·9% (95% CI 79·5–89·0) for culture-confirmed
Abdomen 9 (2%) ··
tuberculosis and 81·4% (76·6–85·3) for all tuberculosis
Pericardium 6 (2%) ··
diagnoses, giving a negative predictive value of 84·6%
Brain 6 (2%) ··
(80·6–87·9) and negative likelihood value of 0·22
Musculoskeletal system 5 (1%) ··
(0·17–0·27) for all tuberculosis cases. Specificity was
Chest wall 2 (1%) ··
86·2% (82·3–89·4) for all patients without tuberculosis
Other 31 (9%) ··
and 93·5% (86·6–97·0) for cases with no risk factors for
Pneumonia ·· 104 (24%)
latent tuberculosis infection (category 4D, table 4).
Sarcoidosis ·· 38 (9%)
Sensitivity of QFT-GIT was 70·6% (64·4–76·1) for
Cancer ·· 36 (8%) culture-confirmed tuberculosis and 67·3% (62·0–72·1)
Lower respiratory tract infection ·· 23 (5%) for all tuberculosis cases, giving a negative predictive
Reactive lymphadenopathy ·· 18 (4%) value of 74·0% (69·5–78·0) and negative likelihood ratio
Chest infection ·· 16 (4%) of 0·41 (0·35–0·48) for all tuberculosis cases (table 4).
Exacerbation of asthma ·· 14 (3%) Specificity was 80·4% (76·1–84·1) for all patients without
Upper respiratory tract infection ·· 13 (3%) tuberculosis and 93·4% (86·4–96·9) for cases with no
Non-tuberculosis mycobacterium infection ·· 12 (3%) risk factors for latent tuberculosis infection. Sensitivity
Exacerbation of bronchiectasis ·· 11 (3%) and specificity of T-SPOT.TB were superior to QFT-GIT;
Exacerbation of chronic obstructive ·· 8 (2%) the relative sensitivity was 1·20 (95% CI 1·12–1·29,
pulmonary disease p<0·0001) and the relative specificity was 1·07 (1·02–1·12,
Other‡ ·· 158 (36%) p=0·0039).
*Some patients had multiple diagnoses. †Some patients had tuberculosis at multiple anatomical sites. ‡Less than Second-generation and ESAT-6-free IGRA results were
five cases per diagnosis. available for 809 (96%) of 845 patients. Sensitivity of
second-generation IGRA was 94·0% (95% CI 90·0–96·4)
Table 3: Final diagnoses of patients with and without active tuberculosis
for culture-confirmed tuberculosis and 89·2%

198 www.thelancet.com/infection Vol 19 February 2019

 Articles

T-SPOT.TB QFT-GIT ESAT, CFP-10, and Rv3615c CFP-10, Rv3615c, and Rv3879c
n/N Estimate (95% CI) n/N Estimate (95% CI) n/N Estimate (95% CI) n/N Estimate (95% CI)
Sensitivity for active tuberculosis
All 253/311 81·4% (76·6–85·3) 220/327 67·3% (62·0–72·1) 273/306 89·2% (85·2–92·2) 263/299 88·0% (83·8–91·2)
Culture-confirmed tuberculosis 185/218 84·9% (79·5–89·0) 163/231 70·6% (64·4–76·1) 203/216 94·0% (90·0–96·4) 197/211 93·4% (89·2–96·0)
Highly probable tuberculosis* 68/93 73·1% (63·3–81·1) 57/96 59·4% (49·4–68·7) 70/90 77·8% (68·2–85·1) 66/88 75·0% (65·0–82·9)
Smear-positive tuberculosis† 45/55 81·8% (69·7–89·8) 42/56 75·0% (62·3–84·5) 48/51 94·1% (84·1–98·0) 47/50 94·0% (83·8–97·9)
Smear-negative tuberculosis†‡ 169/206 82·0% (76·2–86·7) 148/222 66·7% (60·2–72·5) 183/207 88·4% (83·3–92·1) 176/202 87·1% (81·8–91·1)
Pulmonary tuberculosis 79/105 75·2% (66·2–82·5) 79/115 68·7% (59·7–76·5) 88/100 88·0% (80·2–93·0) 85/97 87·6% (79·6–92·8)
Extrapulmonary tuberculosis 141/169 83·4% (77·1–88·3) 113/171 66·1% (58·7–72·8) 148/167 88·6% (82·9–92·6) 142/164 86·6% (80·5–91·0)
Specificity for active tuberculosis
Active tuberculosis excluded 319/370 86·2% (82·3–89·4) 304/378 80·4% (76·1–84·1) 296/370 80·0% (75·6–83·8) 296/372 79·6% (75·2–83·4)
Active tuberculosis excluded, 87/93 93·5% (86·6–97·0) 85/91 93·4% (86·4–96·9) 84/92 91·3% (83·8–95·5) 84/93 90·3% (82·6–94·8)
TST negative, no risk factors for LTBI
Predictive values for all tuberculosis
Positive predictive value 253/304 83·2% (78·6–87·0) 220/294 74·8% (69·6–79·5) 273/347 78·7% (74·1–82·7) 263/339 77·6% (72·8–81·7)
Negative predictive value 319/377 84·6% (80·6–87·9) 304/411 74·0% (69·5–78·0) 296/329 90·0% (86·2–92·8) 296/332 89·2% (85·4–92·1)
Likelihood ratios for all tuberculosis
Positive likelihood ratio ·· 5·90 (4·55–7·66) ·· 3·44 (2·76–4·27) ·· 4·46 (3·62–5·49) ·· 4·31 (3·51–5·28)
Negative likelihood ratio ·· 0·22 (0·17–0·27) ·· 0·41 (0·35–0·48) ·· 0·13 (0·10–0·19) ·· 0·15 (0·11–0·21)

25 of 845 QuantiFERON-TB Gold In-Tube (QFT-GIT) tests, 36 of 845 T-SPOT.TB tests, and 36 second-generation IGRAs were missing because of blood draw difficulties, samples being unsuitable for testing, or
samples being destroyed for laboratory reasons; missing results were spread across all diagnostic categories. Indeterminate and borderline IGRA results were excluded from the analysis and are not presented in
this table; numbers of indeterminate and borderline results for T-SPOT.TB or QFT-GIT and second-generation IGRA are presented in the appendix. When indeterminate and borderline results were included as test
positives in sensitivity analyses (ie, the result could not exclude a tuberculosis diagnosis), sensitivity results for all tuberculosis cases were: 83·2% (95% CI 78·9–86·8) for T-SPOT.TB, 69·7% (64·7–74·2) for QFT-GIT,
90·4% (86·9–93·1) for second-generation IGRA (ESAT-6, CFP-10, and Rv3615c), and 89·6% (85·9–92·4) for ESAT-6-free IGRA (CFP-10, Rv3615c, and Rv3879c). IGRAs=interferon-γ release assays. LTBI=latent
tuberculosis infection. TST=tuberculin skin test. *Highly probable tuberculosis includes culture-negative tuberculosis cases plus ten patients with a final diagnosis of tuberculosis who were not tested for
Mycobacterium tuberculosis; sensitivity (95% CIs) results for culture-negative tuberculosis alone were: 69·9% (95% CI 59·3–78·7) for T-SPOT.TB, 57·1% (46·5–67·2) for QFT-GIT, 75·0% (64·5–83·2) for
second-generation IGRA (ESAT-6, CFP-10, and Rv3615c), and 73·1% (62·3–81·7) for ESAT-6-free IGRA (CFP-10, Rv3615c, and Rv3879c). †Smear microscopy not done for 56 of 845. ‡Among 165 patients who
were smear-negative but culture-positive, 122 of 142 participants were T-SPOT.TB-positive; 105 of 153 were QFT-GIT-positive; 135 of 144 were positive in second-generation IGRA, and 131 of 141 were positive
in ESAT-6-free IGRA.

Table 4: Diagnostic accuracy of commercially available and second-generation IGRAs for diagnosis of active tuberculosis

(85·2–92·2) for all tuberculosis cases, giving a negative Of the 232 patients with culture-confirmed tuberculosis
predictive value of 90·0% (86·2–92·8) and negative and with available smear microscopy results, 165 (71%)
likelihood ratio of 0·13 (0·10–0·19) for all tuberculosis had smear-negative results (57 [35%] of 165 patients
cases. Specificity was 80·0% (75·6–83·8) for all patients had pulmonary tuberculosis, 80 [48%] patients had
without tuberculosis and 91·3% (83·8–95·5) for cases extrapulmonary tuberculosis, and 28 [17%] patients had
with no risk factors for latent tuberculosis infection. both). Sensitivities of the tests in this population were:
Sensitivity of ESAT-free IGRA was 93·4% (89·2–96·0) 85·9% (95% CI 79·2–90·7) for T-SPOT.TB, 68·6%
for culture-confirmed tuberculosis and 88·0% (60·9–75·4) for QFT-GIT, 93·8% (88·5–96·7) for second-
(83·8–91·2) for all tuberculosis cases, giving a negative generation IGRA, and 92·9% (87·4–96·1) for ESAT-6-free
predictive value of 89·2% (85·4–92·1) and negative IGRA.
likelihood ratio of 0·15 (0·11–0·21) for all tuberculosis Of the 135 study participants with HIV, 25 (19%) had a
cases. Specificity was 79·6% (75·2–83·4) for patients final diagnosis of active tuberculosis and 108 (80%) had a
without tuberculosis and 90·3% (82·6–94·8) for cases diagnosis that excluded tuberculosis. 27 (31%) of
with no risk factors for latent tuberculosis infection. 88 participants with diabetes had a final diagnosis of
Comparing second-generation IGRA with T-SPOT.TB, tuberculosis (table 2). Sensitivity and specificity of all
relative sensitivity was 1·08 (95% CI 1·04–1·11, p<0·0001) IGRAs for active tuberculosis in patients with HIV and
and relative specificity was 0·94 (0·91–0·96, p<0·0001). diabetes are shown in the appendix.
For ESAT-6-free IGRA versus T-SPOT.TB, relative The proportion of indeterminate test results did not
sensitivity was 1·07 (1·03–1·10, p=0·0002) and relative differ between QFT-GIT (79 [10%] of 820 patients) and
specificity was 0·93 (0·90–0·96, p<0·0001). A cross- TSPOT.TB (57 [7%] of 809 patients, p=0·061), but was
tabulation of second-generation IGRA against T-SPOT. higher for QFT-GIT than for second-generation IGRA
TB results and table of response magnitudes for each (55 [7%] of 809 patients, p=0·038) and ESAT-6-free IGRA
individual antigen are provided in the appendix. (55 [7%] of 809 patients, p=0·038). Most indeterminate
Responses to Rv3615c were similar to or higher than results occurred in patients without tuberculosis
those to ESAT-6 and CFP-10. (appendix). T-SPOT.TB results were borderline in

www.thelancet.com/infection Vol 19 February 2019 199

 Articles

17 (5%) of 345 patients with active tuberculosis and have the required diagnostic sensitivity to fulfil this role.
16 (4%) of 423 of patients with a non-tuberculosis Although Xpert MTB/RIF Ultra has shown diagnostic
diagnosis. Lowering the cutoff of T-SPOT.TB from eight to sensitivity of 88%, its sensitivity in smear-negative,
five spot-forming cells (thereby scoring all borderline culture-positive tuberculosis is only 63%3 (and sensitivity
results as positive) did not improve diagnostic performance of Xpert MTB/RIF was even lower), 4 compared with
of T-SPOT.TB or either of the second-generation IGRAs, 93·8% (95% CI 88·5–96·7) for second-generation IGRA
giving only a marginal increase in sensitivity at the cost of in this diagnostically challenging subgroup who
a decrease in specificity. Scoring both indeterminate and frequently have paucibacillary disease. However, the very
borderline results as positives also did not affect test high specificity of molecular tests, such as Xpert, provides
performance in sensitivity analyses (table 4). high positive predictive value, enabling rule in of active
tuberculosis. Second-generation IGRA might, thus, have
Discussion a complementary role to rapid molecular tests in the
To our knowledge, this study is the largest prospective diagnostic investigation of suspected tuberculosis.
cohort study embedded in routine clinical practice to Given that IGRAs are the standard of care for detecting
assess and compare the role of IGRAs in the evaluation latent tuberculosis infection,10,11 they will inevitably
of suspected pulmonary and extrapulmonary tuberculosis identify latent tuberculosis infection when active
in a low-incidence setting. Although T-SPOT.TB had tuberculosis has been excluded. Because most people
significantly higher sensitivity than QFT-GIT, neither with possible tuberculosis in low-burden countries are
assay had sufficient sensitivity or negative predictive from ethnic groups with a high prevalence of latent
value to rule out a diagnosis of active tuberculosis. By tuberculosis infection,22 similar to our study, the
contrast, the second-generation IGRA, incorporating diagnostic specificity for active tuberculosis is low for all
Rv3615c with ESAT-6 and CFP-10, had significantly IGRAs, and would be lower still in high-burden
higher diagnostic sensitivity than T-SPOT.TB and countries. The enhanced diagnostic sensitivity of the
QFT-GIT. Also reflecting common clinical practice second-generation IGRA was accompanied by only a
(despite the absence of good evidence or guidelines modest reduction in specificity to 80%, similar to
supporting use of IGRAs in this setting), IGRAs were QFT-GIT. Our study confirms that the low specificity and
used as part of routine diagnostic investigation for active positive likelihood ratios of available and second-
tuberculosis in 35% of study patients; these patients were generation IGRAs mean that a positive result cannot rule
distributed across the recruiting sites (data not shown). in a diagnosis of tuberculosis. However, the specificity of
The negative likelihood ratio of 0·13 for second- all IGRAs increased to 90–93% in patients with active
generation IGRA means a negative test result would tuberculosis excluded and no risk factors for latent
reduce the odds of tuberculosis post-test by a clinically tuberculosis infection (category 4D). Thus, a positive
meaningful factor of 7·7 times compared with pretest IGRA result might help to keep a diagnosis of active
(appendix). The negative predictive value for all tuberculosis in the differential diagnosis in populations
tuberculosis cases, including highly probable cases, was with a very low prevalence of latent tuberculosis infection,
90%, despite the 43% prevalence in this population which, however, is not usually the case in patient
presenting to urban infectious diseases and respiratory populations being assessed for possible tuberculosis.
medicine services with suspected tuberculosis. Since our Two of the leading new tuberculosis vaccine candidates,
study was done in routine clinical practice and Hybrid 1-IC3123 and H56:IC31,24 contain ESAT-6 and
encompassed the full, natural clinical spectrum of might induce conversion of IGRA results in vaccinated
tuberculosis and non-tuberculosis diagnoses, the results individuals. If these vaccines show protective efficacy
are probably generalisable across clinical practice in high- in ongoing clinical trials and achieve licensure,
income, low-incidence countries. Accordingly, in clinical ESAT-6-containing IGRAs will give false-positive results
settings with a low-to-moderate pretest probability of in vaccinated people who are not M tuberculosis infected,
tuberculosis, such as general medical inpatient and analogous to false-positive tuberculosis skin test results
outpatient services or primary care, second-generation in M tuberculosis-uninfected people with previous BCG
IGRA has sufficiently low negative likelihood ratio to vaccination. Diagnostic accuracy of ESAT-6-free IGRA
almost rule out tuberculosis. For example, a negative test was very similar to second-generation IGRA and,
result would convert pretest probability of 20% to a post- therefore, has the potential to replace other IGRAs in
test probability of 3·1%, and a 10% pretest probability to populations immunised against tuberculosis with
1·4% post-test probability (appendix). This test would ESAT-6-based vaccines.
therefore provide a useful prompt triage of patients on Two of the most important global risk factors for
initial presentation, similar to the role other diagnostic tuberculosis are HIV co-infection25 and diabetes,26 both
tests of high sensitivity and restricted specificity have, of which have been reported to adversely affect
such as serum D-dimer in patients with low-to-moderate the performance of IGRAs.27,28 The performance of
suspicion of venous thromboembolism.21 To our commercially available IGRAs in patients with HIV and
knowledge, other available tests for tuberculosis do not diabetes in this study was insufficient to be of value in

200 www.thelancet.com/infection Vol 19 February 2019

 Articles

the diagnosis of active tuberculosis. Performance seemed Contributors

to be lower in HIV-infected and diabetic subgroups, but HSW was responsible for the daily management of the Interferon-γ
Release Assays (IGRA) for Diagnostic Evaluation of Active Tuberculosis
the small numbers of patients with tuberculosis in these (IDEA) study group, including oversight of clinical and laboratory data
subgroups precluded statistical comparisons, which was collection and management. AB managed participant recruitment,
also the case for other types of immunosuppression follow-up activities, and clinical data collection, and contributed to data
associated with tuberculosis, such as chronic kidney management. AAB contributed to laboratory data collection, data
management, and quality assurance. YT did the statistical analyses,
disease and immunosuppressive medication. produced tables and figures, and contributed to data interpretation.
Strengths of our study include the rigorous case MR-R led the study set-up and initial management, and built the study
definitions, including 6-month follow-up to confirm that databases. CP contributed to statistical analyses and producing tables
a diagnosis of tuberculosis was excluded when a non- and figures. HL contributed to laboratory data collection and managed
the laboratory database. JI led the expert clinical panel. GC, ML, CC,
tuberculosis diagnosis was not made at presentation. For DM, FC, FAP, MW, and GW contributed to patient recruitment, data
highly probable tuberculosis, we used a composite collection, and the study expert diagnostic clinical panel. JJD contributed
reference standard1 that was applied by a panel of experts to study design, data analyses, and interpretation of results. OMK and
and experienced clinicians, masked to IGRA results. AL co-led study conceptualisation, design, oversight, and interpretation
of results. Writing of the manuscript was co-led by HSW and AL, and all
Despite this stringent case definition, a proportion of authors contributed to its drafting and revision.
patients without tuberculosis were likely to have been
IDEA study group
incorrectly categorised as having highly probable David Adeboyeku, Robert Davidson (Northwest London Hospitals
tuberculosis, which would explain why all IGRAs had National Health Service [NHS] Trust, London); Martin Dedicoat
lower sensitivity for highly probable tuberculosis than for (University Hospitals Birmingham Foundation Trust, Birmingham);
all tuberculosis cases, including culture-confirmed cases. Heinke Kunst (Queen Mary University Blizard Institute, London);
Michael R Loebingher, Anton Pozniak (Chelsea and Westminster
Therefore, our estimates of diagnostic sensitivity for Hospital NHS Foundation Trust, London); William Lynn (Ealing NHS
all tuberculosis cases, including highly probable Trust, Southall); Nazim Nathani (Sandwell and West Birmingham
tuberculosis, are likely to be conservative, which Hospitals NHS Trust, Birmingham); Rebecca O’ Connell (Bart’s
Healthcare NHS Trust, London); Sarah Menzies (Frimley Health NHS
highlights the significance of increased IGRA sensitivity
Foundation Trust, Camberley).
in culture-confirmed tuberculosis (and the importance of
Declaration of interests
including this subgroup in study analyses) since this is
HSW, AB, AAB, MR-R, and HL were employed by Imperial College
the only population in whom tuberculosis diagnoses are London, London, UK, with grant funding from the National Institute for
definitive. Health Research (NIHR) to do the work described in this Article. JJD,
Our study also has some limitations. First, it does not YT and CP, while working for the University of Birmingham, Birmingham,
UK, received grant funding from NIHR to do the work described in this
include children, in whom the unmet clinical need for
Article. JJD reports grants from NIHR during the study, outside the
improved diagnosis of tuberculosis is high. Second, the submitted work. OMK is employed by Imperial Healthcare Trust and was
numbers of patients with risk factors associated with partly paid by the NIHR grant from Imperial College London. OMK
immunosuppression that affect (eg, HIV) or might affect received other grants from NIHR during the conduct of the study and has
received speaker fees from Oxford Immunotec. He chairs a
(eg, diabetes) test performance were modest, precluding non-remunerated independent committee that organises an annual
clear conclusions about test performance in these educational symposium on tuberculosis, sponsored by Qiagen. AL is
subpopulations. Finally, although blood collection and named inventor on patents (EP05729257.5, EP1735623[B1], US8,105,797[B2],
assays were done strictly in accordance with manufacturers’ EP2069792, EP2069792[B1], EP2005182, EP2005182[B1], US8,765,336[B2],
EP10716313.1, EP2417456[B1], US9,377,460[B2], US9360480[B2],
instructions, IGRAs were not done in a routine diagnostic EP0941478[B2], EP1152012[B1], EP1735623[B1], US8105797[B2],
service laboratory, and retesting of new samples was not EP1144447[B1], and US9005902[B2]) pertaining to T-cell-based diagnosis,
done for cases when initial results were indeterminate or including current and second-generation IGRA technologies. Some of
borderline (as recommended by manufacturers). these patents were assigned by the University of Oxford, Oxford, UK, to
Oxford Immunotec plc, resulting in royalty entitlements for the University
Although the QFT-GIT has been replaced by the of Oxford and AL. All other authors declare no competing interests.
QFT-GIT-Plus since our study was done, its diagnostic
Acknowledgments
accuracy does not appear to be significantly better than We thank the research nurses at each of the study sites for their work in
that of QFT-GIT and no evidence is available to suggest it recruitment of participants; the research assistants at the Tuberculosis
is as sensitive as T-SPOT.TB.29,30 Therefore, our conclusion Research Centre, in particular Bianca Bartlett, Ann-Kathrin Reuschl, and
that neither existing IGRA has a clinically useful role in Luigi Marongiu, who contributed to processing of participant samples
and did the study assays; and Alice Halliday who contributed to the data
the evaluation of suspected active tuberculosis is analysis. We also thank the members of the study steering committee,
unaffected by availability of QFT-GIT-Plus. Khalid Khan, Stephen Gordon, James Gray, Johannes B Reitsma, and
In summary, our study provides conclusive and Nisha Karnani (lay member), for their support and advice throughout
generalisable evidence that existing IGRAs do not have a the study. We acknowledge the support of the National Institute for
Health Research (NIHR) Imperial Biomedical Research Centre and the
useful role as rule-in or rule-out tests for tuberculosis in NIHR Health Protection Research Unit in Respiratory Infections.
routine clinical practice. However, second-generation
References
IGRAs have a higher sensitivity and negative predictive 1 Dosanjh DP, Hinks TS, Innes JA, et al. Improved diagnostic
value, which might help to rule out a diagnosis of evaluation of suspected tuberculosis. Ann Intern Med 2008;
148: 325–36.
tuberculosis in clinical settings with a low-to-moderate
prevalence of tuberculosis.

www.thelancet.com/infection Vol 19 February 2019 201

 Articles

2 Whitworth HS, Aranday-Cortes, Lalvani A. Biomarkers of 17 Alonzo TA, Pepe MS, Moskowitz CS. Sample size calculations for
tuberculosis: a research roadmap. Biomark Med 2013; 7: 349–62. comparative studies of medical tests for detecting presence of
3 Dorman SE, Schumacher SG, Alland D, et al. Xpert MTB/RIF Ultra disease. Stat Med 2002; 21: 835–52.
for detection of Mycobacterium tuberculosis and rifampicin 18 Wilson EB. Probable inference, the law of succession, and statistical
resistance: a prospective multicentre diagnostic accuracy study. inference. J Amer Stat Assoc 1927; 22: 209–12.
Lancet Infect Dis 2018; 18: 76–84. 19 Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial
4 Sohn H, Aero AD, Menzies D, et al. Xpert MTB/RIF testing in a proportion. Stat Sci 2001; 16: 101–17.
low tuberculosis incidence, high-resource setting: limitations in 20 Simel DL, Samsa GP, Matchar DB. Likelihood ratios with
accuracy and clinical impact. Clin Infect Dis 2014; 58: 970–76. confidence: sample size estimation for diagnostic test studies.
5 Garcìa-Basterio AL, Cobelens F. Triage tests: a new priority for J Clin Epidemiol 1991; 44: 763–70.
tuberculosis diagnostics. Lancet 2015; 3: 177–78. 21 Van der Hulle T, Cheung WY, Kooij S, et al. Simplified diagnostic
6 Whitworth HS, Scott M, Connell DW, Donges B, Lalvani A. management of suspected pulmonary embolism (the YEARS
IGRAs–the gateway to T cell based TB diagnosis. Methods 2013; study): a prospective, multicentre cohort study. Lancet 2017;
61: 52–62. 390: 289–97.
7 Pai M, Zwerling A, Menzies D. Systematic review: T-cell-based 22 Pareek M, Greenaway C, Noori T, Munoz J, Zenner D. The impact
assays for the diagnosis of latent tuberculosis infection: an update. of migration on tuberculosis epidemiology and control in
Ann Intern Med 2008; 149: 177–84. high-income countries: a review. BMC Med 2016; 14: 48.
8 Goletti D, Carrara S, Butera O, et al. Accuracy of immunodiagnostic 23 Mearns H, Geldenhuys HD, Kagina BM, et al. H1:IC31 vaccination
tests for active tuberculosis using single and combined results: is safe and induces long-lived TNF-α+IL-2+CD4 T cell responses in
a multicenter TBNET-Study. PLoS One 2008; 3: e3417. M. tuberculosis infected and uninfected adolescents: a randomised
9 Sester M, Sotgiu G, Lange C, et al. Interferon-γ release assays for trial. Vaccine 2017; 35: 132–41.
the diagnosis of active tuberculosis: a systematic review and 24 Luabeya AK, Kagina BM, Tameris MD, et al. First-in-human trial of
meta-analysis. Eur Respir J 2011; 37: 100–11. the post-exposure tuberculosis vaccine H56:IC31 in Mycobacterium
10 UK International Clinical Guidelines Team. Tuberculosis: tuberculosis infected and non-infected healthy adults. Vaccine 2015;
prevention, diagnosis, management and service organisation. 33: 4130–40.
London: National Institute for Health and Care Excellence, 2016. 25 Esmail H, Riou C, Bruyn ED, et al. The immune response to
11 ECDC. Use of interferon-gamma release assays in support of TB Mycobacterium tuberculosis in HIV-1-coinfected persons.
diagnosis. Stockholm: European Centre for Disease Prevention and Annu Rev Immunol 2018; 36: 603–38.
Control; 2011. 26 Kuo MC, Lin SH, Lin CH, Mao IC, Chang SJ, Hsieh MC.
12 Du F, Xie L, Zhang Y, et al. Prospective comparison of QFT-GIT Type 2 diabetes: an independent risk factor for tuberculosis:
and T-SPOT.TB assays for diagnosis of active tuberculosis. Sci Rep a nationwide population-based study. PLoS One 2013; 8: e78924.
2018; 8: 5882. 27 Cattamanchi A, Smith R, Steingart KR, et al. Interferon-gamma
13 Liu XQ, Dosanjh D, Varia H, et al. Evaluation of T cell responses to release assays for the diagnosis of latent tuberculosis infection in
novel RD1 and RD2 encoded Mycobacterium tuberculosis gene HIV-infected individuals: a systematic review and meta-analysis.
products for specific detection of human tuberculosis infection. J Acquir Immune Defic Syndr 2011; 56: 230–38.
Infect Immun 2004; 72: 2574–81. 28 Faurholt-Jepson D, Aabye MG, Jenson AV, et al. Diabetes is
14 Millington KA, Fortune SM, Low J, et al. Rv3615c is a highly associated with lower tuberculosis antigen-specific interferon
immunodominant RD1 (Region of Difference 1)-dependent gamma release in Tanzanian tuberculosis patients and
secreted antigen specific for Mycobacterium tuberculosis infection. non-tuberculosis controls. Scand J Infect Dis 2014; 46: 384–91.
Proc Natl Acad Sci USA 2011; 108: 5730–35. 29 Takasaki J, Manabe T, Morino E, et al. Sensitivity and specificity of
15 Ruhwald M, de Thurah L, Kuchaka D, et al. Introducing the ESAT-6 QuantiFERON-TB Gold Plus compared to QuantiFERON-TB Gold
free IGRA, a companion diagnostic for TB vaccines based on In-Tube and T-SPOT.TB on active tuberculosis in Japan.
ESAT-6. Sci Rep 2017; 7: 45969. J Infect Chemother 2018; 24: 188–92.
16 Li G, Li F, Zhao HM, et al. Evaluation of a new IFN-γ release assay 30 Barcellini L, Borroni E, Brown J, et al. First independent evaluation
for rapid diagnosis of active tuberculosis in a high incidence of QuantiFERON-TB Plus performance. Eur Respir J 2016;
setting. Front Cell Infect Microbiol 2017; 7: 117. 47: 1587–90.

202 www.thelancet.com/infection Vol 19 February 2019