Article Text
PDF
Abstract
Background The value of interferon-γ release assays (IGRA) to diagnose active tuberculosis (TB) in children is not established, but these assays are being widely used for this purpose. The authors examined the sensitivity of commercially available IGRA to diagnose active TB in children in the UK compared with the tuberculin skin test (TST).
Methods The authors established a paediatric tuberculosis network and conducted a retrospective analysis of data from children investigated for active TB at six large UK paediatric centres. All centres had used TST and at least one of the commercially available IGRA (T-Spot.TB or Quantiferon-Gold in Tube) in the diagnostic work-up for active TB. Data were available from 333 children aged 2 months to 16 years. The authors measured the sensitivity of TST and IGRA in definite (culture confirmed) and probable TB in children, agreement between TST and either IGRA, and their combined sensitivity.
Results Of 333 children, 49 fulfilled the criteria of definite TB, and 146 had probable TB. Within the definite cohort, TST had a sensitivity of 82%, Quantiferon-Gold in tube (QFT-IT) had a sensitivity of 78% and T-Spot.TB of 66%. Neither IGRA performed significantly better than a TST with a cut-off of 15 mm. Combining the results of TST and IGRA increased the sensitivity to 96% for TST plus T-Spot.TB and 91% for TST plus QFG-IT in the definite TB cohort.
Conclusions A negative IGRA does not exclude active TB disease, but a combination of TST and IGRA increases the sensitivity for identifying children with active TB.
Statistics from Altmetric.com
Introduction
There is now a large body of evidence supporting the use of commercially available interferon-γ (IFNγ) release assays (IGRA) to aid the diagnosis of tuberculosis (TB) infection in a variety of settings. Two types of IGRA are currently available in the UK: T-Spot.TB (Oxford Immunotec, Oxford, UK) and Quantiferon-TB Gold In-Tube (Cellestis, Victoria, Australia). Both assays work on the principle that the T cells of the individuals who have acquired TB infection will respond to restimulation with TB antigens by secreting IFNγ. T-Spot.TB is an ELISPOT assay. Peripheral blood mononuclear cells (PBMC) are isolated and then stimulated with the TB-specific antigens ESAT-6 and CFP10. Antigen-specific cells can subsequently be quantified by enumerating the individual cells which have produced IFNγ.1 Quantiferon-TB Gold In-Tube (QFG-IT) uses a similar principle but is an in-tube assay. Whole blood is stimulated with ESAT-6, CFP10 and an additional TB-specific antigen TB7.7. Supernatants are then harvested, and secreted IFNγ quantified using ELISA.2
What is already known about this topic
Interferon-γ release assays (IGRA) can distinguish a positive tuberculin skin test (TST) due to BCG vaccination or environmental non-typical mycobacterial (NTM) exposure from a positive TST due to infection with Mycobacterium tuberculosis.
However, the IGRA cannot differentiate between active and latent tuberculosis (TB).
What this study adds
Acknowledging that a positive IGRA is not designed to discriminate between active or latent TB, it can nevertheless provide confirmatory information in children who are being investigated for active TB, especially if used in conjunction with TST.
However, a negative IGRA does not exclude active TB.
A series of reviews and meta-analyses3,–,7 concluded that IGRA have a high sensitivity and specificity, particularly in the diagnosis of latent TB infection (LTBI) in immune competent adults. However, based on limited evidence, it appears that IGRA may be less sensitive in the diagnosis of active TB infection, particularly in children.3,–,10
In 2006, the UK National Institute for Health and Clinical Excellence (NICE) recommended the use of IGRA in diagnosing latent TB in adults and children.11 No recommendations were made regarding employment of IGRA for diagnosis of active TB. However, in the UK and internationally, IGRA have been employed for this purpose. IGRA were welcomed as a new tool to improve the difficult diagnosis of active TB in children.
We established a paediatric UK TB network (PTBNET-UK) to accumulate data on a large number of culture-confirmed and probable cases of childhood TB to investigate further the sensitivity of two IGRA in the UK setting. Data from six paediatric units treating children with TB across the UK indicate that IGRA fail to detect active TB reliably.
Methods
In October 2007, we hosted a national workshop addressing use of IGRA in paediatrics in the UK. All centres present decided to combine data on diagnosis of paediatric active TB. The analysis of these data is the subject of this report and provides the opportunity to compare performance of IGRA with tuberculin skin test (TST) in a large paediatric cohort.
Participating centres had access to IGRA over the preceding 36–48 months. At St Mary's Hospital, this formed part of a dedicated research project comparing both IGRA (T-Spot.TB; Oxford Immunotec; QFG-IT; Cellestis) side by side with TST.12 At Newcastle NHS Trust, initially only Quantiferon assays were performed via routine laboratory service. Later, T-Spot.TB was added within a comparative IGRA/TST research project. At other centres, IGRA were carried out via routine laboratories, and usually only one IGRA was available, depending on individual NHS-Trust laboratory practice. One centre used T-Spot.TB only, two used QFG-IT only, and three used both assays (figure 1). All centres used TST at the time of assessment, and IGRA were performed regardless of TST size.
Distribution of cases (probable and definite) from each centre including numbers of children tested with QFT, T.Spot-TB or both tests (GOSH, Great Ormond St Hospital; NHG, Newcastle General Hospital; RHSC, Glasgow Royal Hospital for Sick Children; RLCH, Royal Liverpool Children's Hospital; RLH, Royal London Hospital; SMH/NWP, Northwick Park/St Mary's Hospital).
Inclusion criteria
Children between 2 months and 16 years old attending NHS services for management of presumed active TB either as inpatients or as outpatients between 1 January 2005 and 31 December 2007 were included in the analysis.
Exclusion criteria
Children with possible TB were excluded from analysis. Possible TB was defined as signs and symptoms potentially consistent with active TB and existing risk factors for active TB such as a history of TB contact or history of travel to TB endemic countries, but the treating paediatrician subsequently made an alternative diagnosis or felt that active TB was unlikely. None of the children in this group received treatment for active TB, and none developed active TB to date.
Definition of study groups
Cases of active TB were defined according to two categories:
Definite TB: children with culture confirmed TB.
Probable TB: absence of culture confirmation, but all children fulfilled all of the following criteria: (A) clinical symptoms and signs of active TB, (B) abnormal radiography consistent with TB and/or abnormal cerebrospinal fluid (CSF) consistent with tuberculous meningitis, (C) response to TB therapy plus, (D) either a history of TB contact or travel to a TB-endemic country within the last 24 months.
Microbiology
Children admitted with suspected active pulmonary TB had sputum samples and/or gastric washings collected. Children with suspected TB meningitis had CSF microscopy and culture. For other sites of presumed disseminated disease, tissue biopsies were obtained where possible and appropriate.
Tuberculin skin testing
Two units of purified protein derivative (PPD) RT23 (Staten Serum Institute, Denmark) were injected to the volar surface of the left forearm according to the intradermal Mantoux method by the medical or TB nursing staff. The transverse diameter (in millimetres) of skin induration was recorded 48–72 h later.
Laboratory assays
In all institutions except St Mary's Hospital, samples were sent to routine NHS laboratories. At St Mary's Hospital, all samples were processed in the research laboratories of the principal investigator by a technician with no access to patient information. T-Spot.TB and QFG-IT assays were carried out according to the manufacturer's recommendations, as described previously.12
Statistical analysis
A data-collection tool was developed at the IGRA workshop. Participating centres forwarded data to the Medical Research Council-Clinical Trials Unit (MRC-CTU) for analysis. Assay sensitivity was defined as the proportion of positive results identified within the case and definite cohort. Indeterminate results/test failures were not excluded in calculating sensitivity estimates. Agreement between assays was measured by total percentage agreement and kappa (κ) statistic.13 Strength of agreement was defined as poor (κ≤0.2), fair (κ=0.21–0.4), moderate (κ=0.41–0.6), good (κ=0.61–0.8) and very good (κ=0.8–1). Analyses were performed using STATA, version 10 (StataCorp 2007, Stata Statistical Software: Release 10; StataCorp, College Station, Texas).
Results
Data from 333 children were compiled. Figure 1 summarises the participating sites, numbers of children enrolled and type of assays conducted at individual sites. TST was carried out at all sites.
Of 333 children, 195 fulfilled the criteria for definite or probable TB (case cohort) and were commenced on full TB treatment. Forty-nine children had culture-confirmed TB (definite cohort), and 146 children were classified as probable TB (probable cohort). The remaining 138 children were not finally diagnosed as TB. Either an alternative diagnosis was made or the treating paediatrician did not find enough evidence to support a final TB diagnosis. Their data were not included for further analysis. Demographic and clinical details for the case cohort are shown in table 1.
Demographic details, BCG status and tuberculosis (TB) contact history for case cohort
For analysis purposes, results are summarised per individual test for the whole case cohort (definite and probable TB) as well as separately for the ‘gold standard’ definite TB cases and according to TST induration above 15 mm. This cut-off is in line with the recommendations by NICE for employing IGRA in diagnosis of TB infection in BCG vaccinated children and is used by UK clinicians as a cut-off also in active disease.11 14
TST results
Of 195 children in the case cohort, 172 had TST. Of these, 108 (63%) had TST>15 mm. Of 49 children in the definite cohort, 45 had TST. Of these, 37 (82%) had TST>15 mm. The sensitivity of TST>15 mm in confirmed active TB was 82% (95% CI 68% to 92%). No adverse events were reported following administration of TST.
T-Spot.TB results
Of 195 children in the case cohort, T-Spot.TB results were available for 103 children, of which 47 (46%) were positive. Of 49 children in the definite cohort, T-Spot.TB results were available for 27 children, of which 18 (67%) were positive. The sensitivity of T-Spot.TB in confirmed active TB was 67% (95% CI 46% to 83%). Of the 108 children from the case cohort with a TST>15 mm, T-Spot.TB results were available for 50 children, of which 34 (68%) were positive.
Nine (9%) of the T-Spot.TB tests gave no result through technical failure. The mean age at testing of children with failed T-Spot.TB was 4.7 years, which was significantly lower than the mean age at testing of the rest of the case cohort tested with T-Spot.TB (p=0.017).
QFG-IT results
Of 195 children in the case cohort, QFG-IT results were available for 170 children, of which 101 (59%) were positive. Of 49 children in the definite cohort, QFG-IT results were available for 46 children, of which 36 (78%) were positive. The sensitivity of QFG-IT in confirmed active TB was 78% (95% CI 64% to 89%). Of 108 children from the case cohort with a TST>15 mm, QFG-IT results were available for 97 children, of which 79 (81%) were positive.
Thirteen (8%) of the QFG-IT test results were indeterminate. The mean age at testing of these children was 10.2 years, which was not significantly different from the mean age of 8.5 years of the case cohort tested with QFG-IT (p=0.243). The results are summarised in tables 2, 3.
Results of T-Spot.TB testing and Quantiferon-TB Gold In-Tube (QFG-IT) testing for the case cohort, definite cohort and those with tuberculin skin test (TST)>15 mm
Results of tuberculin skin test (TST) for case cohort and definite cohort
Agreement between T-Spot.TB and TST
In the case cohort, 91 children had results available for both TST and T-Spot.TB. The percentage agreement between the two tests was 69.1% (κ=0.37, SE=0.11, fair agreement). There was no significant difference between the sensitivity of the two tests (p=0.56). In the definite cohort, 25 children had results available for both TST and T-Spot.TB. The percentage agreement between tests was 58.3% (κ=−0.15, SE=0.15, poor agreement).
Agreement between QFG-IT and TST
In the case cohort, 158 children had results available for TST and QFG-IT. The percentage agreement between tests was 78.5% (κ=0.54, SE=0.08, moderate agreement). There was no significant difference between sensitivities of the two tests (p=0.86). In the definite cohort, 43 children had available results for both TST and QFG-IT. The percentage agreement between tests was 85% (κ=0.41, SE=0.16, moderate agreement). The results are summarised in table 4.
Results of T-Spot.TB and QFG-IT compared with result of tuberculin skin test (TST) for both the definite and the case cohort
Combining TST and IGRA
Combining a TST>15 mm and T-Spot.TB (defined as either one or both are positive) correctly identified 24/25 definite TB cases (sensitivity 96%). Combining TST>15 mm with QFG-IT identified 39/43 cases of definite TB (sensitivity 91%). When conducting the same analysis for the entire case cohort, TST plus T-Spot.TB identified 61/91 (67%), and TST plus QFG-IT identified 114/158 (72%) of all cases.
Discussion
The utility of IGRA in the investigation of active TB, especially in children, remains a matter of debate.5,–,7 15 This study is the first to compare IGRA and TST results from a large paediatric cohort with culture confirmed TB in the UK. We have shown that IGRA may have limited sensitivity in diagnosing active TB and that it is no more sensitive than a TST>15 mm. However, combining both TST and IGRA resulted in correct identification of over 90% of all culture-confirmed cases. This study demonstrates that larger patient numbers can be collected by establishing a network of paediatricians caring for TB patients in different parts of the country sharing a common interest and data platform. PTBNET-UK is now in a position to conduct prospective studies of paediatric TB.
A recently updated meta-analysis summarised currently available data on test sensitivity in active TB in adults and children.6 Pooled data from adults and children gave a sensitivity of 90% (95% CI 86% to 93%) for T-Spot.TB, 78% (95% CI 73% to 82%) for QFT-2G (an older version of the Quantiferon-system) and 70% (95% CI 63% to 78%) for QFG-IT. However, this meta-analysis included only two studies involving substantial numbers of children. The authors concluded that the sensitivity of IGRAs may be somewhat lower in children than in adults.
In line with our own findings, Domínguez et al16 reported a prospective comparison of QFG-IT with T-Spot.TB. Only nine children included in the study were diagnosed as having active TB. It is not reported whether these were culture-confirmed. Six out of nine (66.7%) had positive T-Spot.TB, 6/9 (66.7%) had positive QFG-IT, and 9/9 (100%) had positive TST. Detjen et al17 also reported the highest sensitivity of the TST (100%, 95% CI 88% to 100%), as well as 93% sensitivity (95% CI 77% to 99%) for QFG-IT and 93% sensitivity (95% CI 77% to 99%) for T-Spot.TB in a head-to-head comparison of all three tests. The cut-off value for TST was 5 mm, which probably increased its sensitivity at the price of specificity. More recently, Bianchi et al18 found a sensitivity of 93.8% for QFG-IT and 43.8% for TST>15 mm in active TB, but the sample size was small.
Why IGRA appear to perform less well in children than adults is an important question. It is well known that young children are highly susceptible to TB and often present with more severe disease. Just as in vivo immune responses to TB in children appear to be less effective than those in adults, in vitro responses to TB specific antigens also appear less robust.19 Differences in the dynamics of the immune response to TB associated with an increased susceptibility to active TB and the development of severe disease in young children have been summarised by Lewinsohn et al,20–21 and Kampmann et al22 have previously demonstrated a lower production of IFNγ in response to mitogen stimulation in children younger than 4 years compared with older children. Connell et al23 have similarly demonstrated a positive correlation between age and IFNγ production, although Lalvani and Millington have reported no difference in mitogen response with age using the ELISPOT technique (apart from during the neonatal period).15 It is likely that low levels of TB antigen-specific IFNγ secreting T cells in addition to more generalised T cell immune suppression observed in active disease might lead to impaired responses in IGRA, especially in children. Our results support this hypothesis, but we acknowledge that other smaller studies have demonstrated a better sensitivity.17 23,–,26
Our study has a number of limitations. It was not powered to examine the effect of age, sex, BCG status, country of origin or pattern of disease on IGRA results. Such data could be gathered by increasing the network research in this area to collect data on a larger number of patients.
IGRA were carried out only at the time of diagnosis. Longitudinal data assessing potential change over time and with treatment might be informative, for example, in the context of poor clinical response, as recently proposed by Herrmann et al.27
We are unable to estimate specificity for each test or the positive predictive value. This is a common situation in assessment of IGRA, especially in children. Owing to the paucibacillary nature of active disease in children and the tendency to avoid more invasive diagnostic techniques such as bronchoscopy, it is difficult to prove that a child definitely does not have active TB. The reported study population also has relatively high rates of latent TB infection, which further complicates this issue. It is therefore not possible to measure accurately the number of false positives or true negatives, which are necessary to calculate either of these parameters. Larger prospective national studies will potentially address this issue.
Similar to results from other investigators, we have noted discrepancies between IGRA tests.16 28,–,30 We found differences in sensitivity between QFG-IT and T-Spot.TB approaching statistical significance (p=0.09), with QFG-IT being more sensitive than T-Spot.TB. These findings contrast with the published meta-analysis of predominantly adult populations, which found T-Spot.TB (in low incidence countries) to be more sensitive than TST and previous versions of the QFG-IT test in active culture confirmed TB6 and highlighted the potentially different read-outs of the immune response using whole blood or PBMC.31 32
The study has limited power to detect significant differences between IGRA test using the McNemar test, and the observed non-significant p values could be a reflection of the relatively small numbers. Increasing numbers in future studies should address this problem.
Higher rates of indeterminate results/test failures have been reported in paediatric samples compared with adults.15 30 In our study, rates of indeterminate results were similar for QFG-IT and T-Spot.TB at approximately 10%, in line with Connell et al26 who found no significant difference between rates of indeterminate results. Other studies have suggested that QFG might give indeterminate results more frequently than T-Spot.TB in children.15 The mean age of children included in this study is 8.5 years. A larger study including a greater number of younger children would be of use to investigate this potential issue further. It should also be born in mind that different generations of QFG assays exist, and different techniques are used for the collection and preparation of PBMC. Definitive conclusions cannot necessarily be drawn between data generated using different protocols.
There are proposed advantages of IGRA over TST. They are laboratory-based and therefore more easily standardised than the TST, they require only one patient contact, and there appears to be no risk of affecting future testing through a booster effect, unlike TST. There may also be an advantage in immunosuppressed states such as HIV infection where IGRA appear to be more sensitive than TST.24 32 These potential benefits should be considered, and further data from paediatric populations are needed to complete the picture.
The proportion of positive results for any of the tests in our study was approximately 20% higher in the cohort of children with definite TB compared with the entire case cohort (combined definite and probable cases). This either reflects an even lower sensitivity of IGRA in culture-negative active TB in children or could be an indicator of overdiagnosis and unnecessary treatment of children mistakenly classified as having active TB. Treatment for TB in children is more often initiated on clinical grounds than on microbiological confirmation. We recommend that any validation of novel diagnostic tests in children should be carried out for both proven ‘gold-standard’ definite TB and probable TB in order to reflect any additional diagnostic value that can be expected from the use of IGRA or other novel diagnostic test. A more detailed study, examining results of IGRA and TST in relation to level of probability of infection and response to treatment in culture-negative probable TB, would be required to elucidate further.
How useful are IGRA in the diagnosis of active TB in children?
Although we have shown that IGRA may have suboptimal sensitivity, this does not rule out an additional beneficial role for TB diagnosis in children. When combining results for TST and either IGRA, over 90% of children with culture-confirmed tuberculosis were correctly identified. Sensitivity decreased to 67–72% when probable TB was also included. However, this figure probably represents a more realistic estimate of the sensitivity in daily practice, since microbiological confirmation is the exception rather than the rule.
If it can be assumed that the specificity of IGRAs in active TB is similar to that in LTBI, a positive IGRA in a child with signs and symptoms suggestive of active TB infection in a region with low TB prevalence is strongly supportive of a diagnosis of active TB. However, our results suggest that a negative IGRA in the same situation should not be used to rule out TB infection.
In areas of high TB prevalence with high rates of LTBI, a positive IGRA in a child with signs suggestive of active TB may not be as useful. Presence of LTBI leads to a positive IGRA, which may be incidental to the disease causing the symptoms or signs under investigation. In this situation, a test that accurately distinguishes active from latent TB infection would be of much higher value.
In conclusion, a negative IGRA does not exclude TB, but IGRA may be a useful adjunct in the diagnostic pathway for active TB in children in the UK. Decision to treat, however, should continue to include clinical findings, imaging and risk factor analysis in addition to microbiological investigations, even if those are known to have a relatively poor yield. The PTBNET-UK will be useful to conduct prospective multicentre studies and to investigate current and novel diagnostic, therapeutic and preventive approaches to paediatric tuberculosis.
References
Footnotes
-
Funding Consumable grants from the PEEL Foundation (The Peel Medical Research Trust, Sceptre Court, 40 Tower Hill, London EC3N 4DX, UK), the European Society for Paediatric Infectious Diseases (c/o Kenes International, Rue de Chantepoulet 1–3, PO Box 1726, CH-1211 Geneva 1, Switzerland) and the HPA Mycobacterium Reference Unit (Clinical Sciences Research Centre, Centre for Infectious Disease, Institute of Cell and Molecular Science, Barts and the London Queen Mary's School of Medicine and Dentistry, 2 Newark Street, London E1 2AT, UK) contributed to this work. BK is funded by the Wellcome Trust. She is supported by the Biomedical Research Centre and is a coinvestigator at the Centre for Respiratory Infection (CRI) at Imperial College London. STA is supported by a European Union grant (European Commission, EuropeAid Co-operation Office, B 1049 Brussels, Belgium): aid for poverty-related diseases in developing countries (EuropeAid/121404/C/G/Multi). AB is funded by the NIHR (National Institute of Health Research, Room 132, Richmond House, 79 Whitehall, London SW1A2NL, UK) as a Training Fellow. None of the funding sources had a role in the design of the study, data collection, analysis, data interpretation or writing of the paper.
Competing interests None.
Ethical approval Study approval was obtained from the ethics committees of Imperial College Healthcare NHS Trusts, North West London Hospitals Trust and Newcastle NHS Trust.
Provenance and peer review Not commissioned; not externally peer reviewed.
Patient consent Obtained from the parents.