Evaluation of QuantiFERON-TB Gold Plus for Predicting Incident Tuberculosis among Recent Contacts: A Prospective Cohort Study.

1. Institute for Global Health, University College London, London, UK 2. Blizard Institute, Queen Mary University of London, London, UK 3. UCL-TB and UCL Respiratory, University College London, London, UK 4. Royal Free London NHS Foundation Trust, London, UK 5. Division of Infection & Immunity, University College London, UK 6. TB Unit, Public Health England, Colindale, London, UK 7. North Middlesex University Hospital NHS Trust, London, UK


Screening for latent tuberculosis infection (LTBI) among recent tuberculosis (TB) contacts is an
important component of TB control, particularly in settings with low TB incidence aiming towards pre-elimination (1). However, currently available LTBI diagnostics lack sensitivity, and have poor predictive value for incident TB (2)(3)(4)(5)(6)(7)(8) Consequently, prevention of one incident TB case requires treatment of many for LTBI. This is true for both interferon gamma release assays (IGRAs) and the tuberculin skin test (TST); a recent evaluation found that QuantiFERON Gold-In-Tube (QFT-GIT; Qiagen, Hilden, Germany) and T-SPOT.TB (Oxford Immunotec, UK) perform similarly to the TST when a BCG-stratified TST cut-off is used (8).
A newer generation QuantiFERON (QuantiFERON-TB Gold Plus; QFT-Plus) was recently launched, adding a second TB antigen tube (TB2) that incorporates short peptides designed to stimulate a CD8 + T-cell response, in addition to the CD4 + -response tube (TB1) included in previous versions. The proposed rationale for this is that CD8 + -responses have been associated with mycobacterial load and recent TB exposure (9,10). Initial independent evaluations have suggested QFT-Plus may have improved test sensitivity in active TB compared to QFT-GIT (11), and that the CD8 + -targeted antigen tube response may be associated with proxy measures of degree of TB exposure among contacts (12). However, no studies have examined the prognostic value of QFT-Plus for predicting incident TB. We aimed to address this key knowledge gap in a prospective cohort of UK TB contacts.
Participants completed a questionnaire and had blood sampling for QFT-Plus (at least six weeks from last known TB exposure). Contacts with evidence of prevalent TB disease (defined as TB diagnosed within 21 days of enrolment as per previous work(8)), and those accepting preventative therapy (offered in accordance with contemporary national guidance (13,14)) were excluded from analysis.
Participants were linked to national TB surveillance records held by Public Health England, including all statutory TB notifications, to identify those notified with TB (until 31/12/2017). Notified TB cases were validated by local record review and included cultureconfirmed TB or clinically diagnosed with radiological or histological evidence of TB, where a clinician had prescribed treatment with a full course of anti-TB treatment.
QFT-Plus results were interpreted as per manufacturer guidance, with TB antigen responses calculated as TB antigen interferon-γ minus unstimulated control interferon-γ. We calculated incidence rates and rate ratios relative to the negative test category, along with sensitivity, specificity and predictive values, including the full duration of follow-up.
To assess the incremental value of adding the CD8 + -stimulating tube in predicting incident TB cases, we compared receiver operating characteristic (ROC) curves and areas under the curves (AUCs) when using TB1 only, TB2 only, and the maximal TB antigen tube (higher of TB1 and TB2). We also plotted a ROC curve for the calculated difference between the TB1 and TB2 tubes (TB2-TB1) as a surrogate for the CD8 + -specific response, since it has been hypothesised that this may identify contacts with recently acquired M. tuberculosis infection, who are at highest risk of TB disease (12).

Results
A total of 623 contacts were recruited, of whom 532 (85.4%) had QFT-Plus results (89 missing; 2 indeterminate) and were followed-up for a median 1.93 years (interquartile range (IQR) 1.65-2.21). QFT-Plus results were positive in 180/532 (33.8%) (Table 1), of whom 39 (21.7%) commenced preventative therapy. One patient was notified with prevalent TB (3 days after recruitment). A total of 492 participants were therefore included in the analysis. Baseline characteristics were similar between included and excluded participants, except for those who commenced preventative therapy being younger than those who did not (Table 1).

Discussion
We found that QFT-Plus performance appeared comparable to published evaluations of QFT-GIT and T-SPOT.TB, with an IRR of 10.1, 80% sensitivity for detection of incident TB, and an overall PPV for incident TB of 5.7% (8). Interferon-γ responses in the TB1 and TB2 tubes were strongly correlated, and ROC curves showed minimal difference between them for predicting incident TB. As a result, the calculated difference between TB1 and TB2 responses, as a proxy for the CD8-specific response, did not predict incident TB. However, despite our sample size of 492 recent TB contacts, the number of TB progressors was small, reflecting low progression risk even among contacts. Thus, a larger-scale study is indicated to investigate subtle differences in the relative prognostic contributions of the TB1 and TB2 antigen tubes. This is the first evaluation of the prognostic value of QFT-Plus test. The prospective design allowed the collection of detailed clinical, demographic and laboratory data. Participants were recruited while attending routine contact-tracing services, ensuring the study population was representative of TB contacts. Our findings are therefore likely generalizable to other low incidence settings globally. Moreover, follow-up was robust through linkage to national surveillance records using a validated matching algorithm (15), minimising risk of missing incident TB cases.
A limitation was that the provision of preventative therapy to a subset of the QFT-Plus positive patients could have led to selection bias. However, while the subgroup who received preventative therapy were younger than those who did not (reflecting national policy at the time of the study (13,14)), other characteristics were similar, suggesting the impact of this bias was likely small. Secondly, the TB contacts included both pulmonary and extra-pulmonary index cases, reflecting national contact screening policy during the study period (13). PPV of QFT-Plus may be higher among populations including only pulmonary TB contacts, due to higher pre-test probability of incident TB. However, previous evaluations of the QFT-GIT and T-SPOT.TB have also included extra-pulmonary contacts, which allows the current study findings to be put into this context (8). Thirdly, serial testing (before and after exposure) was not performed, meaning that we are unable to assess QFT-Plus conversions over time, which may provide a more reliable measure of recent M. tuberculosis infection. This reflects the reality of contact screening practices; the ability of assays to accurately stratify TB risk from a single baseline test is therefore a key attribute. The absence of serial testing also means that participants who developed incident TB may have been re-exposed to M. tuberculosis during the interval between testing and disease, though the overall risk of exposure in the UK (a low TB incidence setting) is likely small. Fourthly, QFT-Plus results were missing or indeterminate for 91/623 patients (14.6%), in keeping with the proportion of missing results for other IGRAs in our recent evaluation (8). However, these patients' characteristics were similar to the overall study population, suggesting that risk of subsequent selection bias was likely small. Finally, we included both microbiologically-confirmed and clinically-diagnosed TB cases in our outcome definition, in keeping with previous IGRA evaluations (3)(4)(5)(6)(7)(8). The rationale for this is that extra-pulmonary TB, which occurs frequently among TB cases occurring in foreign-born people living in the UK(16), is often challenging to prove microbiologically. However, all TB cases diagnosed during the study received a full course of TB therapy, and none were de-notified. It is therefore likely that these represented true TB cases, with low risk of outcome misclassification.
In summary, in this first evaluation of the predictive value of QFT-Plus for incident TB, performance was comparable to other commercial IGRAs. Better biomarkers are required to transform management of TB contacts.