Comparison of T-SPOT.TB Assay and Tuberculin Skin Test for the Evaluation of Young Children at High Risk for Tuberculosis in a Community Setting
OBJECTIVE. We wished to compare the sensitivity of an enzyme-linked immunospot assay (T-SPOT.TB; Oxford Immunotec, Oxford, United Kingdom) and the tuberculin skin test for the detection of tuberculosis infection in very young children being evaluated for active tuberculosis in a rural community setting.
METHODS. Children with a history of exposure to tuberculosis and children presenting to a local clinic or hospital with symptoms suggesting tuberculosis were admitted to a dedicated case verification ward. T-SPOT.TB testing was performed, and children were evaluated with a clinical examination, a tuberculin skin test, chest radiographs, and cultures of induced sputum and gastric lavage specimens. The diagnosis was determined by using a clinical algorithm.
RESULTS. A total of 243 children (median age: 18 months) were recruited, of whom 214 (88%) had interpretable T-SPOT.TB results. Children ≥12 months of age were more likely than younger children to have positive T-SPOT.TB results, whereas tuberculin skin test results were unaffected by age. The sensitivity of the T-SPOT.TB was no better than that of the tuberculin skin test for culture-confirmed tuberculosis (50% and 80%, respectively) and was poorer for the combined group of culture-confirmed and clinically probable tuberculosis (40% and 52%, respectively). For the 50 children clinically categorized as not having tuberculosis, the specificity of both the T-SPOT.TB and the tuberculin skin test was 84%.
CONCLUSIONS. For young children presenting in a community setting after exposure to tuberculosis or with symptoms suggesting tuberculosis, T-SPOT.TB cannot be used to exclude active disease. The sensitivity of this assay may be impaired for very young children.
Tuberculosis in children remains a neglected area of research despite considerable morbidity and mortality rates, particularly in sub-Saharan Africa.1 The diagnosis is challenging because of the difficulty of obtaining appropriate specimens, paucibacillary disease,2 and nonspecific clinical presentation.3 Culture remains the standard method but lacks sensitivity.4 Therefore, there has been interest in the potential utility of interferon γ release assays, such as the enzyme-linked immunospot assay (ELISpot), for the diagnosis of active and latent tuberculosis infection in children.5–11 Because the antigens used in these assays are not present in BCG or most environmental mycobacteria, they are proposed to offer improved specificity over the tuberculin skin test (TST), which uses a crude extract of proteins from Mycobacterium tuberculosis.7
Neither an interferon γ release assay nor the TST is able to distinguish between active disease and latent infection, and a positive result for a child with suspected tuberculosis is nonspecific for active disease, particularly in high-incidence areas. However, a highly sensitive marker would be useful in excluding infection in children who present with suspected tuberculosis.
Three prospective studies6,8,9 evaluated ELISpot for the diagnosis of active childhood tuberculosis and documented sensitivities ranging from 81% to 93%. Those studies were predominately hospital-based, which suggests that more-ill patients were enrolled. It is possible that ELISpot may be more sensitive among children with less-severe disease, because interferon γ responses were shown previously to be depressed in severe illness in adults.12 We wished to determine the sensitivity of ELISpot for active childhood tuberculosis in a community setting.
This study was performed in the context of a phase IV trial conducted by the South African Tuberculosis Vaccine Initiative, which compared percutaneous and intradermal BCG vaccination (A.H., M.H., F. Little, M. A. Goetz, L. Barker, H. Mahomed, J. Sadoff, W.A.H., L. Geiter, G.H., South African BCG Trial Team, unpublished data, 2008). The study was based in Worcester, South Africa, which has a stable semirural population with a high burden of tuberculosis (smear-positive case detection rate of 602 cases per 100000 in 2004) and a HIV prevalence of 8.1% among prenatal clinic attendees in 2005.
During the course of the study, 11680 BCG-vaccinated children were monitored from birth for ≥2 years. The following categories of children were admitted to a dedicated case verification ward at the regional tuberculosis hospital: (1) children who presented to any hospital or community clinic in the region with symptoms suggesting tuberculosis (loss of weight or failure to gain weight, unexplained fever, or prolonged cough), (2) children who had chest radiographs performed because of respiratory symptoms at any referral hospital, and (3) children who lived at the same address as any adult with tuberculosis diagnosed during the period. All children who presented to the case verification ward during an 11-month period (February to December 2005) were included in the study. Written informed consent was obtained from parents or guardians of all participants, and the study was approved by the University of Cape Town Research Ethics Committee (protocol 271/2000).
The degree of exposure to tuberculosis was quantified on the basis of a history of contact with a patient with active tuberculosis. The nonexclusive exposure categories were as follows: no contact, any contact with a patient with active tuberculosis, contact with an adult currently receiving antituberculous therapy, and contact with a household member with tuberculosis.
All children were thoroughly investigated for tuberculosis, including clinical examination, chest radiographs, acid-fast microscopy and mycobacterial culture of 2 gastric aspirates and 2 induced sputum samples, and a Mantoux skin test (2 tuberculin units of purified protein derivative RT-23 [Statens Serum Institut, Copenhagen, Denmark], equivalent to 5 tuberculin units of purified protein derivative test, read at 48 hours). A final diagnostic classification was assigned by using a diagnostic algorithm (which did not include either TST or ELISpot results) drawn up by an expert clinical panel (Table 1). Chest radiographs were reviewed independently by 3 expert reviewers who were blinded to clinical details, with a minimum of two thirds agreement required. According to the algorithm, children were classified as having definite, probable, or possible tuberculosis or not having tuberculosis. All children were tested with a HIV enzyme-linked immunosorbent assay (ELISA), except when parental consent was refused. Positive ELISA results for children <18 months of age were confirmed with a HIV polymerase chain reaction assay, whereas results for older children were confirmed with a second ELISA. The decision to initiate antituberculous chemotherapy for a child was made by the physician responsible, who was blinded to the results of the ELISpot. Children who were diagnosed as not having tuberculosis but who were in contact with a patient with tuberculosis were referred for isoniazid chemoprophylaxis according to national guidelines. Follow-up care for children diagnosed as not having tuberculosis was through routine clinical services, which are of a high standard in the study area. The South African Tuberculosis Vaccine Initiative maintains ongoing surveillance of all regional clinic and hospital records for pediatric tuberculosis and therefore would have detected readmission of any of the study participants.
The commercial T-SPOT.TB assay (Oxford Immunotec, Oxford, United Kingdom) was used. At the time of admission, 3 to 4 mL of blood were collected into cell preparation tubes (Becton Dickinson, Franklin Lakes, NJ) through peripheral venipuncture, for completion of the assay according to the manufacturer's guidelines. Tubes were transported to the laboratory in insulated boxes and processed if received within 5 hours after collection. Spots were enumerated by 2 observers with a stereomicroscope, and the mean spot count was used for interpretation (interobserver agreement was excellent; correlation coefficient: 0.998). Positive T-SPOT.TB results were reported if either panel of peptides yielded positive results, following the manufacturer's recommendations. T-SPOT.TB results were excluded if samples were received >5 hours after collection, if the positive control assay failed, or if there were insufficient cells to perform the assay.
The primary outcomes were the T-SPOT.TB results and the size of the TST reaction in relation to the assigned clinical categorization. For the Mantoux test, a transverse diameter of induration of ≥10 mm was considered positive. The χ2 test for trend was used to test for associations between the clinical classification or degree of exposure and T-SPOT.TB and TST results. McNemar's test with continuity correction was used for comparisons of proportions, with confidence intervals (CIs) calculated by using the approximation method. The Mann-Whitney test was used to compare ages between clinical categories, as well as times to sample processing for T-SPOT.TB testing. The χ2 test was used to compare the proportions of positive T-SPOT.TB results according to the time to sample processing. Fisher's exact test was used to determine the association between TST or T-SPOT.TB results and the degree of exposure. The κ statistic was used to summarize agreement. The Spearman test was used to determine the correlation between sizes of TST results or numbers of spots and age.
Two hundred forty-three BCG-vaccinated children were evaluated, of whom 214 had interpretable T-SPOT.TB results (Fig 1). The median age of the children was 18 months (interquartile range: 14–24 months) (Table 2). The distribution of TST responses is shown in Fig 2. HIV results were available for 204 (95%) of 214 children. One child (22 months of age) had a positive HIV ELISA result and was included in the analysis (the child had probable tuberculosis, with negative TST and T-SPOT.TB results). None of the children was readmitted with suspected tuberculosis in the study region after discharge from the case verification ward.
T-SPOT.TB and TST Results and Clinical Diagnosis of Active Tuberculosis
Both positive T-SPOT.TB and positive TST responses were strongly associated with increasing likelihood of tuberculosis, as determined with the clinical algorithm (P = .0023 for T-SPOT.TB and P < .0001 for TST) (Fig 3A). T-SPOT.TB and TST responses were positive for 50% (95% CI: 24%–76%) and 80% (95% CI: 49%–94%) of children with culture-confirmed tuberculosis, respectively (P = .25). When the combined group of 58 children with definite or probable tuberculosis was considered, T-SPOT.TB was less sensitive (sensitivity: 40% [95% CI: 27%–53%]) than TST (sensitivity: 52% [95% CI: 38%–65%]; P = .046). The specificities of the 2 tests, when evaluated by using the clinical category of not having tuberculosis, were identical (specificity: 84% [95% CI: 71%–93%]). The numbers of children with positive TST and T-SPOT.TB results in each diagnostic category are detailed in Fig 4. Of note, only 2 of the 113 children classified as having definite, probable, or possible tuberculosis had negative TST results but positive T-SPOT.TB results. Overall agreement between TST and T-SPOT.TB results was moderate (κ = 0.548 [95% CI: 0.43–0.67]) but agreement was highest among probable cases (definite: κ = 0.4 [95% CI: −0.17 to 0.97]; probable: κ = 0.75 [95% CI: 0.55–0.94]; possible: κ = 0.51 [95% CI: 0.33–0.68]; not tuberculosis: κ = 0.26 [95% CI: −0.16 to 0.67]).
To determine whether the use of T-SPOT.TB or TST would have substantially affected the diagnostic categorization of children, we compared the number of children in each diagnostic category when either TST or T-SPOT.TB was added as a potential additional feature to the diagnostic algorithm (Table 1). There was excellent agreement in categorization of children by the algorithm when either TST or T-SPOT.TB was included as an additional feature (agreement: 93%; κ = 0.902; P < .00001). The inclusion of either T-SPOT.TB or TST results in the diagnostic algorithm would have resulted in reclassification of 7 children (from not tuberculosis to possible tuberculosis in each case); however, not all of the same children were reclassified by using the 2 tests, and inclusion of a positive result on either test would have resulted in 11 children being reclassified. Neither test would have altered the diagnostic categorization of any of the children classified as having probable (or definite) tuberculosis. There was no significant association between anthropometric measures (weight-for-age or height-for-age z scores) and positive T-SPOT.TB results among children with definite or probable tuberculosis (data not shown).
Association Between Degree of Exposure to a Case of Active Tuberculosis and Positive T-SPOT.TB or TST Results
There was increasing likelihood for positive T-SPOT.TB and TST results with increasing exposure (T-SPOT.TB: P = .0003; TST: P = .0081) (Fig 3B). Similar numbers of children with no history of exposure had positive TST and T-SPOT.TB results (14 of 57 children and 11 of 57 children, respectively; P = .51). A history of any contact with a patient with tuberculosis was associated with increased odds of positive TST results (odds ratio [OR]: 2.4 [95% CI: 1.2–4.8]) but not positive T-SPOT.TB results (OR: 1.6 [95% CI: 0.7–3.3]). Both positive TST results and positive T-SPOT.TB results were significantly associated with having a household contact with tuberculosis (OR: 2.4 for both [95% CI: 1.4–4.2 for TST and 1.3–4.6 for T-SPOT.TB]).
Association Between Age and Positive T-SPOT.TB or TST Results
Young children (≤12 months of age) were more likely to have positive TST results (40% [95% CI: 25%–58%]) than positive T-SPOT.TB results (3% [95% CI: <0.0001%–18%]). When only children with a history of exposure to tuberculosis were considered, age did not affect TST responses but did affect T-SPOT.TB responses (children ≥12 months of age were no more likely than younger children to have positive TST results; OR: 0.83 [95% CI: 0.34–2.02]) but were more likely to have positive T-SPOT.TB results (OR: 10.04 [95% CI: 1.31–77.05]).
The proportion of children with positive T-SPOT.TB results increased with age (χ2 test for trend, P < .0001), whereas there was no such trend with TST results (P = .4). There was a weak but significant correlation between age and the total number of spot-forming cells in the T-SPOT.TB (Spearman ρ = 0.22; P = .0065) but not between age and TST diameter (Spearman ρ = 0.13; P = .1).
Effect of Time to Sample Processing on Results
The time between collection and processing of samples did not affect T-SPOT.TB results, with no significant difference in the proportions of positive results when findings were categorized according to the number of hours to processing (P = .5276). The median times to processing for T-SPOT.TB positive (3.5 hours) and negative (3.6 hours) results were similar (P = .319). Furthermore, there was no difference in median times to processing between assays with failed positive control tests (3.11 hours) and those without (3.4 hours; P = .15).
The most important findings of this study were that negative T-SPOT.TB results could not be used to exclude active tuberculosis in young children at risk (because a significant proportion of clinically diagnosed or microbiologically proven cases had negative test results) and that the sensitivity of the T-SPOT.TB but not the TST seemed to be affected by very young age. T-SPOT.TB detected only 5 of 10 culture-proven cases (compared with 8 of 10 for the TST) and 40% of definite or probable cases (compared with 52% for the TST). In contrast, Liebeschuetz et al8 described a sensitivity of 85% for ELISpot and 70% for TST among HIV-negative children admitted to the hospital with confirmed or highly probable tuberculosis at another study site in South Africa. A study conducted in a hospital in a country with low prevalence reported a sensitivity of 93% for ELISpot for the diagnosis of tuberculosis in children with established, culture-proven disease.6 Although overdiagnosis of tuberculosis among probable cases is a possibility, the proportions of positive T-SPOT.TB responses in the definite (50%) and probable (37.5%) categories were similar, which suggests that T-SPOT.TB misses a substantial proportion of true cases in both categories.
The inclusion of either T-SPOT.TB or TST (or both) in the diagnostic algorithm used in this study would not have changed the diagnostic categorization for any of the children categorized as having definite or probable tuberculosis. Only 2 of 164 children classified as having definite, probable, or possible tuberculosis had negative TST results and positive T-SPOT.TB results; therefore, the use of T-SPOT.TB in place of or in addition to TST would have had minimal impact on the sensitivity of the clinical algorithm.
T-SPOT.TB did not offer any improvement in sensitivity over TST for the diagnosis of active tuberculosis. What are the possible reasons for the relatively poor sensitivity of T-SPOT.TB in this study? Firstly, previous studies largely examined children presenting to a hospital with suspected tuberculosis, whereas we also recruited children who attended a local clinic with symptoms suggesting tuberculosis or with a history of contact with a patient with tuberculosis. Those children might have had milder or earlier illness than in previous studies. It is possible that very early presentation and testing might precede the development of adaptive immune responses (as detected with the T-SPOT.TB assay). However, the TST results were positive in 3 of 5 cases of culture-confirmed tuberculosis with negative T-SPOT.TB results, which suggests that, at least for those 3 children, delayed hypersensitivity was present. This difference between the tests may arise because perinatal BCG vaccination primes a cross-reactive TST response to M tuberculosis in children who subsequently become infected. BCG does not prime a recall response to early secreted antigenic target-6 (ESAT-6) and/or culture filtrate protein 10 in the same way, because they are absent from BCG.
Secondly, we used the commercial T-SPOT.TB assay in this study. We reported previously that an in-house ELISpot using recombinant ESAT-6 and culture filtrate protein 10 proteins had a sensitivity of 83% for the diagnosis of culture-confirmed tuberculosis.9 Another reported prospective study examining children with suspected tuberculosis used both recombinant ESAT-6 and overlapping peptides derived from ESAT-6 and culture filtrate protein 10.8 In contrast, the commercial T-SPOT.TB assay uses only pools of overlapping peptides derived from these proteins. The differences in antigen processing and presentation associated with peptides and recombinant antigens may influence responses; because peptides do not require processing, however, this might favor the commercial assay.
The final explanation relates to the possible age dependency of T-SPOT.TB but not TST responses. The median age of children in this study (18 months) was markedly lower than that in the 3 previous studies (32–39 months). Although Liebeschuetz et al8 did not find an association between age and ELISpot responses in their cohort, this might have been attributable to small numbers in the younger age group (median age for patients with confirmed or highly probable tuberculosis: 50 months; interquartile range: 24–84 months). In our study, there was a clear trend for the proportion of positive T-SPOT.TB responses to increase with age. There was no such association between age and positive TST results; therefore, this is unlikely to represent simply an increase in cumulative exposure to tuberculosis with age. The association was particularly striking when children ≤12 months of age who had been exposed to a tuberculosis case were considered. A number of studies have described an association between age and interferon γ production in response to mitogen and mycobacterial antigens in whole blood, suggesting that age may play a role.5,13,14
It may be argued that the cross-sensitization by BCG was responsible for the relatively large proportion of young children with positive TST results. This is a difficult issue to resolve; however, it seems from the strongly bimodal distribution of TST responses, with few TST responses between 0 and 10 mm, that, if BCG does prime TST responses in this community, then these responses are seldom in the range commonly noted after BCG vaccination of infants.15 Similarly, in neighboring Botswana, the majority of TST results of ≥10 mm were attributed to tuberculosis infection, rather than previous BCG vaccination.16 It seems more plausible that the strongly positive TST responses seen were the result of true exposure to M tuberculosis. In addition, the specificities of T-SPOT.TB and TST were identical (84%) for children classified as not having tuberculosis, which suggests that BCG did not play a major role in priming TST responses.
Although it was not the primary outcome, it is interesting to note that TST and T-SPOT.TB results both correlated with the degree of exposure to tuberculosis. Positive TST (but not T-SPOT.TB) results were associated with a history of any exposure to tuberculosis, whereas both positive TST results and positive T-SPOT.TB results were associated with a history of exposure to a household member with tuberculosis. This may indicate that TST is a more-sensitive marker of tuberculosis infection in young children or that T-SPOT.TB is more specific. This result is in keeping with the findings of Hill et al,17 who concluded that ELISpot was less sensitive than TST for the diagnosis of recent tuberculosis infection among children in Gambia.
We have shown that a substantial proportion of young children with clinical and/or microbiologic evidence of active tuberculosis have negative T-SPOT.TB results at presentation. The test may have impaired sensitivity for very young children, for whom it should not be used to exclude the possibility of active tuberculosis.
This study was supported by the Wellcome Trust (grant 072065 to Dr Nicol and grant 072070 to Dr Wilkinson), the Aeras Global Tuberculosis Vaccine Foundation (Drs Hatherill, Hanekom, and Hussey and Mrs Workman), Bristol Myers Squibb “Secure the Future” initiative (Drs Nicol and Beatty), and the South African Medical Research Council (Dr Nicol).
We thank the participants, their parents, and the South African Tuberculosis Vaccine Initiative ward staff members. We also thank Monique Hanslo for her assistance with data analysis.
- Accepted April 10, 2008.
- Address correspondence to Mark Nicol, FCPath, PhD, Institute of Infectious Disease and Molecular Medicine, Faculty of Health Sciences, University of Cape Town, Room N2.10, Anzio Road, Observatory, 7925, Cape Town, South Africa. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on this Subject
In hospital-based studies, interferon γ release assays (such as T-SPOT.TB) have been shown to have sensitivity of 80% to 90% for the diagnosis of pediatric tuberculosis. Their utility among children with milder disease, presenting to community clinics, is unknown.
What This Study Adds
A substantial proportion of young children with clinical and/or microbiologic evidence of active tuberculosis had negative T-SPOT.TB results. The test may have impaired sensitivity for very young children, for whom it should not be used to exclude active tuberculosis.
- ↵Marais BJ, Gie RP, Hesseling AC, et al. A refined symptom-based approach to diagnose pulmonary tuberculosis in children. Pediatrics.2006;118 (5). Available at: www.pediatrics.org/cgi/content/full/118/5/e1350
- ↵Connell TG, Curtis N, Ranganathan SC, Buttery JP. Performance of a whole blood interferon γ assay for detecting latent infection with Mycobacterium tuberculosis in children. Thorax.2006;61 (7):616– 620
- ↵Detjen AK, Keil T, Roll S, et al. Interferon-γ release assays improve the diagnosis of tuberculosis and nontuberculous mycobacterial disease in children in a country with a low incidence of tuberculosis. Clin Infect Dis.2007;45 (3):322– 328
- ↵Nicol MP, Pienaar D, Wood K, et al. Enzyme-linked immunospot assay responses to early secretory antigenic target 6, culture filtrate protein 10, and purified protein derivative among children with tuberculosis: implications for diagnosis and monitoring of therapy. Clin Infect Dis.2005;40 (9):1301– 1308
- Richeldi L, Ewer K, Losi M, et al. T-cell-based diagnosis of neonatal multidrug-resistant latent tuberculosis infection. Pediatrics.2007;119 (1). Available at: www.pediatrics.org/cgi/content/full/119/1/e1
- ↵Hill PC, Brookes RH, Adetifa IM, et al. Comparison of enzyme-linked immunospot assay and tuberculin skin test in healthy children exposed to Mycobacterium tuberculosis. Pediatrics.2006;117 (5):1542– 1548
- Copyright © 2009 by the American Academy of Pediatrics