OUP user menu

Interferon gamma release assay in diagnosis of pediatric tuberculosis: a meta-analysis

Lin Sun, Jing Xiao, Qing Miao, Wei-xing Feng, Xi-rong Wu, Qing-qin Yin, Wei-wei Jiao, Chen Shen, Fang Liu, Dan Shen, A-dong Shen
DOI: http://dx.doi.org/10.1111/j.1574-695X.2011.00838.x 165-173 First published online: 1 November 2011

Abstract

Although interferon gamma release assays (IGRAs) have been widely used for the diagnosis of latent and active tuberculosis in adults, a relative lack of validation studies in children has led to caution in their clinical interpretation. This meta-analysis systematically evaluated two IGRAs (ELISA and ELISPOT) and the tuberculin skin test (TST). We searched databases (PubMed, MEDLINE, Ovid) between January 2000 and January 2011 using search terms of latent tuberculosis infection or tuberculosis and interferon gamma release assay, or T-SPOT.TB test, or QuantiFERON-TB Gold, or ESAT-6, or CFP-10, and child, or childhood, or pediatrics. We also collected data by performing a manual search of references from relevant articles and communicating with selected authors. The meta-analysis was conducted with random effects models to account for heterogeneity between selected studies. The sensitivities of all three tests in active tuberculosis were similar. The pooled sensitivity was 70% for ELISA studies, 62% for ELISPOT studies and 71% for TST. Calculated sensitivities for IGRAs and the TST differ in culture-confirmed tuberculosis [ELISA (85%) vs. ELISPOT (76%) vs. TST (85%)] and clinical diagnosed cases [ELISA (64%) vs. ELISPOT (58%) vs. TST (66%)]. The pooled specificity was 100% for ELISA and 90% for ELISPOT, but was much lower for TST [56% in all included studies and 49% in children with bacillus Calmette-Guerin (BCG) vaccination]. The agreement between the TST and IGRAs in non-BCG-vaccinated children is higher than that in BCG-vaccinated children. In the diagnosis of active tuberculosis in children, the TST and IGRAs have similar sensitivity. By contrast, the specificity of IGRAs is far greater than the TST, particularly in children with previous BCG vaccination.

Keywords
  • meta-analysis
  • interferon gamma
  • diagnosis
  • tuberculosis
  • children

Introduction

Childhood tuberculosis is commonly extra-pulmonary, disseminated and severe, especially in children under 3 years of age, and is associated with high morbidity and mortality (Marais et al., 2006). In children, diagnosis of tuberculosis is complicated by its pauci-bacillary nature, resulting in atypical clinical signs and a lower probability of bacteriological confirmation (Rigouts, 2009). Currently, the diagnosis of latent tuberculosis infection (LTBI) is hindered by the lack of a ‘gold standard’. The tuberculin skin test (TST) was until recently the main method of detecting Mycobacterium tuberculosis infection and in diagnosing active tuberculosis. The TST uses a poorly defined mix of antigens from M. tuberculosis resulting in false-positive responses because of nontuberculous mycobacteria (NTM) infection or previous bacillus Calmette-Guerin (BCG) vaccination. False-negative TST results can occur when children suffer from severe active tuberculosis or immune suppression.

Therefore, alternative diagnostic tools for the detection of tuberculosis have been explored. The interferon gamma release assays (IGRAs) are based on two antigens: the early-secreted antigenic target 6-kDa protein (ESAT-6) and culture filtrate protein 10 (CFP-10). Several commercially available IGRA tests have been developed to assist in the diagnosis of latent and active M. tuberculosis infection, including the T-SPOT.TB test (TSPOT) (Oxford Immunotec, Oxford, UK) and QuantiFERON-TB Gold (QFT-G) or QuantiFERON-TB Gold In-Tube (QFT-IT) (Cellestis, Carnegie, Australia). TSPOT uses the ELISPOT (enzyme-linked immunosorbent spot) technique to measure the number of individual mycobacterium-specific T cells. QFT-G and QFT-IT measure the concentration of interferon gamma produced in whole blood with enzyme-linked immunosorbent assay (ELISA).

The assays have been widely used for identifying or diagnosing tuberculosis infection and have become useful additional tests in the diagnosis of active tuberculosis in adults. In contrast, only a few studies have reported their utility in children. Therefore, in this meta-analysis, we aimed to compare the sensitivity and specificity of commercial IGRAs with the TST in pediatric tuberculosis.

Materials and methods

Study strategy

We conducted a literature search of databases (PubMed, MEDLINE, Ovid) for articles published between January 2000 and January 2011. Search terms included latent tuberculosis infection or tuberculosis and interferon gamma release assay, or T-SPOT.TB test, or QuantiFERON-TB Gold, or ESAT-6 or CFP-10, and child, or childhood or pediatrics. We performed manual searches of the references from relevant articles and corresponded with the authors of some articles for complete information.

Articles were included if they met the following selection criteria.

(1) Articles that reported original data were included; reviews, case reports and editorials were excluded. Articles with fewer than five enrolled subjects were excluded.

(2) Studies that presented data on the sensitivity and specificity of the commercial versions of IGRAs, including T-SPOT.TB, QuantiFERON-TB Gold or QuantiFERON-TB Gold In-Tube, were included. For studies assessing sensitivity, the participants were required to have active tuberculosis confirmed by bacteriological evidence or diagnosed by clinical evidence. For studies assessing specificity, the participants should be low-risk individuals without identified exposure to active tuberculosis. Participants coinfected with HIV or other immune compromises and those who had received antituberculosis treatment were excluded.

(3) In studies that evaluated the concordance of the tests, all tests should have been done simultaneously and in the same people to ensure comparability.

Two independent reviewers (L.S. and J.X.) performed searches and selected articles according to the inclusion criteria designed in advance. One reviewer abstracted both the test and participant characteristics of the articles collected. A second reviewer double-checked these data.

Statistical analysis

For each study, we calculated sensitivity, specificity, positive rate and 95% confidence intervals (CIs) and summarized the results in forest plots. Studies were weighted by total sample size to pool estimates of sensitivity and specificity across the studies. Statistical analysis was conducted using meta-disc, version 1.4 (Hospital Ramony Cajal, Madrid, Spain). We evaluated heterogeneity by using the chi-square test and I2 test. The random effects model (DerSimonian and Laird) was performed when heterogeneity was present (P < 0.05 and I2 > 50%), and the fixed effects model (Mantel–Haenszel) otherwise.

Results

Eligible studies

After independent review, 16 articles including 598 patients with tuberculosis and 432 controls were available for analysis. Twelve of these studies were published in the last 3 years (Fig. 1). Fourteen studies assessed sensitivity of IGRAs and the TST among children diagnosed with active tuberculosis. Of these, nine studies performed ELISA, and 10 studies performed ELISPOT. Seven studies assessed specificity of IGRAs and the TST among control children without identified tuberculosis exposure. Of these, three studies used ELISA and five studies used ELISPOT (Table 1).

Figure 1

Flow chart of article selection.

View this table:
Table 1

Characteristics of studies included for analysis of sensitivity and specificity

Cut-offSensitivitySpecificity
StudyCountryAge (years)Index testTSPOTQFT-IT/QFT-GTST (mm)TB patients, Bac/Clin (n)Patients (n)BCG vaccinated (%)
Sun et al. (2010)China≤18TSPOT/TST≥6 spots1018/565190.2
Warier et al. (2010)India≤18TSPOT≥6 spots1015/384792.5
Hansted et al. (2009)Lithuania10–17TSPOT/TST≥6 spots1023/052100
Lighter et al. (2009a)USA≤17QFT- IT/TST17.5 pg mL−1107/NR210
Detjen et al. (2007)Germany0.3–15QFT- IT/TSPOT/TST≥6 spots> 0.35 IU mL−1528/0220
Cruz et al. (2011)USA0.1–18TSPOT/TST≥6 or ≥8 spots513/18
Nicol et al. (2009)South Africa> 18TSPOT/TST≥6 spots1010/48
Kampmann et al. (2009)UK0.25–16QFT- IT/TSPOT/TST≥6 spots> 0.35 IU mL−110 and 1525/38
Bamford et al. (2010)UK0.17–16QFT- IT/TSPOT/TST≥6 spots> 0.35 IU mL−115195/0
Grare et al. (2010)France0.5–11QFT- IT/TST> 0.35 IU mL−1150/7
Tsolia et al. (2010)Greece≤15QFT- IT/TST> 0.35 IU mL−110 or 513/12
Bianchi et al. (2009)Italy≤16QFT- IT/TST> 0.35 IU mL−156/10
Lighter et al. (2009b)USA≤18QFT- IT/TST> 0.35 IU mL−11030NR
Domínguez et al. (2008)Spain≤18QFT- IT/TSPOT/TST≥6 spots> 0.35 IU mL−159/0
Connell et al. (2008)Australia0.5–19QFT-G/TSPOT/TST≥6 spots> 0.35 IU mL−159/NR
Soysal et al. (2008)Turkey6–10TSPOT/TST≥6 spots10 and 15209100
  • TSPOT, T-SPOT.TB test; QFT-IT, QuantiFERON-TB Gold In-Tube; QFT-G, QuantiFERON-TB Gold; TST, tuberculosis skin test; Bac, bacteriology; Clin, clinical course; NR, not reported.

Sensitivity of IGRAs and the TST

For studies assessing sensitivity, the active tuberculosis cases were confirmed by culture or by standard clinical criteria. All cases were without HIV infection. Figure 2 shows the separate sensitivities of all three tests in active tuberculosis. For ELISA, nine studies could be included, resulting in a pooled sensitivity of 70% (229/328, 95% CI 65–75%). For ELISPOT, 10 studies were included, resulting in a pooled sensitivity of 62% (277/443, 95% CI 57–67%). For the TST, 12 studies were available, resulting in a pooled sensitivity of 71% (365/512, 95% CI 67–75%). Levels of heterogeneity between the studies were high (I2 = 80.3% for ELISA, 92.3% for ELISPOT, 85.8% for TST).

Figure 2

Forest plot of studies estimating sensitivity of the three tests in patients with active tuberculosis: (a) ELISA, (b) ELISPOT, (c) TST. The red circles and horizontal lines correspond to the recorded percentage of true positive results among tuberculosis cases and their respective 95% CI. The area of the red circles reflects the weight each study contributes to the analysis. The diamond represents the pooled value with its 95% CI. Failed or indeterminate test results were not included in the analysis.

Tuberculosis patients were further divided into two subgroups: culture-confirmed tuberculosis and clinically diagnosed tuberculosis. We observed an increased pooled sensitivity for ELISA (85%), ELISPOT (76%) and TST (85%) in the former subgroup, while the pooled sensitivity dropped distinctly in cases diagnosed clinically, with sensitivity for ELISA (64%), ELISPOT (58%) and TST (66%).

We then divided tuberculosis patients by grouping their countries according to tuberculosis incidence rates. High-incidence-rate countries (≥50 per 100 000) included India, China, South Africa and Lithuania. All acceptable published studies for ELISA were done in low-incidence-rate countries and showed a pooled sensitivity of 70%. Studies on ELISPOT and the TST were conducted in both high- and low-incidence countries and a similar pooled sensitivity was found (ELISPOT: 64 vs. 61%; TST: 71 vs. 71%).

Specificity of IGRAs and the TST

For studies assessing specificity, the participants enrolled were low-risk individuals without identified exposure to active tuberculosis, regardless of BCG vaccination. The rate of BCG vaccination of the participants varied (0–100%). As shown in Fig. 3, the pooled specificity was 100% for ELISA studies (73/73, 95% CI 84–100%), 90% for ELISPOT studies (342/381, 95% CI 86–93%) and 56% for TST studies (214/385, 95% CI 50–61%). Because only three studies targeting the evaluation of ELISA demonstrated a high specificity, we chose to focus on the comparison of ELISPOT and the TST in further analysis. The specificity of ELISPOT was affected only slightly by BCG vaccination status (89% for vaccinated vs. 95% for unvaccinated), or by national incidence status (95% for high-incidence vs. 86% for low-incidence groups). On the other hand, the specificity of the TST was significantly affected by BCG vaccination status (49% for vaccinated vs. 93% for unvaccinated).

Figure 3

Forest plot of studies estimating specificity of the three tests in healthy children without identified exposure to active tuberculosis: (a) ELISA, (b) ELISPOT, (c) TST. The red circles and horizontal lines correspond to the recorded percentage of true positive results among tuberculosis cases and their respective 95% CI. The area of the red circles reflects the weight each study contributes to the analysis. The diamond represents the pooled value with its 95% CI. Failed or indeterminate test results were not included in the analysis.

Concordance between IGRAs and the TST

Seven studies assessed the concordance between IGRAs and the TST with varying rates of BCG vaccination. Among them, five studies concluded that agreement between TST and IGRA tests in non-BCG-vaccinated children is higher than that in BCG-vaccinated children. For example, one study (Tsolia et al., 2010) assessed the concordance between ELISA and the TST according to BCG immunization status, and found that among non-BCG-immunized patients agreement was excellent (κ = 0.34–1.00), while among BCG-immunized children it was fair to poor (κ = 0.02–0.28).

Discussion

It is estimated that pediatric cases account for 10–15% of the global tuberculosis case load. Diagnosis of pediatric tuberculosis is challenging because of the limitations of conventional methods. Culture and microscopy findings are often negative in children. Advances in molecular biology and genomics have led to alternatives to the TST (Pai et al., 2006; Starke, 2006). Commercially available IGRAs have evolved rapidly, and they have been widely used in many settings. Regrettably, researchers have limited access to evaluate the assays in the field of pediatric tuberculosis. Meta-analyses can increase the effective sample size under investigation through the pooling of data from individual association studies, thereby enhancing statistical power for assessing sensitivity and specificity of IGRAs and the TST.

In our review, the sensitivity of the two commercial IGRAs and the TST shows an equivalent sensitivity in active tuberculosis. Results from an earlier review indicated that a lower sensitivity of ELISA and the TST have been found in pediatric tuberculosis compared with adults (Menzies et al., 2007). Our data agree with findings in other studies that the sensitivities of the three tests (62–71%) are lower than those in adults (70–89%) (Jiang et al., 2007; Pai et al., 2008; Diel et al., 2010). The results of the three tests are based on the reaction of the immunological effecter cells. It has been reported that CD4+ T cells were the major cell type producing interferon-gamma (IFN-γ), a type 1 cytokine which plays an important role in the host immune response (Leung et al., 2009). Some authors underline that there were statistically significant differences in CD4+ T-cell subpopulations between children at different ages (P < 0.05) (Lee et al., 1996; Kam et al., 2001). As a consequence, we concluded that a special diagnostic threshold for a positive result may be adjusted for children based on their suboptimal and developmental cellular immune responses.

The relatively poor performance of IGRAs in clinically diagnosed cases remains a concern. Because of the pauci-bacillary nature of the disease, the diagnosis of active tuberculosis is often based on a combination of clinical signs and symptoms, suggestive radiology, history of household exposure, as well as the TST reaction. According to the decreased pooled sensitivity in clinical diagnosed tuberculosis in this analysis, some of the cases could be over diagnosed as active tuberculosis due to the overlap of symptoms with other childhood illnesses. Pediatric tuberculosis clinicians had high hopes that applying the results of IGRAs to guide clinical diagnosis would be more helpful. The disappointing lack of sensitivity of IGRAs in the context of clinical cases may be a result of failure to detect IFN-γ produced by antigen-specific T cells. To date, in the absence of bacterial evidence, we cannot determine if children with active tuberculosis were missing. Large cohort studies are required to elucidate this issue.

The most important finding in this analysis is the significantly low specificity of the TST and a high specificity of IGRAs regardless of BCG vaccination of the subjects enrolled. Although all the children enrolled for assessing specificity have no identified tuberculosis risk, there was a high rate of TST-positive cases — almost all false positive. A significant reason is the effect of BCG vaccination. First, among six studies assessing specificity of the TST, three were available with BCG-vaccinated children, resulting in distinctly low specificities. When these studies were removed from consideration, the pooled specificity of the TST was remarkably improved. The specificity of the IGRAs remained high in mostly BCG-vaccinated children. Our findings were in line with the study of Menzies et al. (2007), emphasizing that the average specificity of IGRAs with RD1 antigens was 97.7 and 92.2% for ELISA and ELISPOT, respectively, and were both more specific than the TST in cases of BCG vaccination. Secondly, when assessing the concordance between IGRAs and the TST with varying rates of BCG vaccination, five studies concluded that agreement between TST and IGRA tests in BCG-vaccinated children is lower than that in non-BCG-vaccinated children. According to our review, BCG immunization can cause false positive reactions in the TST but not in IGRAs which use antigens (ESAT-6 and CFP-10) not present in BCG or in common environmental mycobacteria. The cut-off value for a TST-positive result varies greatly, from 5, 10 and up to 15 mm of induration, and there is no good conclusion supporting a particular reasonable cut-off for injection of PPD (purified protein derivative) intradermally as positive in BCG-immunized children. As a result, despite the cost and complexity of IGRAs, they will be increasingly used in screening LTBI in children with or without identified tuberculosis risk.

Infection with NTM is also associated with high false-positive results. The effect of NTM infection on IGRAs and the TST is poorly studied. Only one study (Detjen et al., 2007) enrolled 23 children with bacteriologically confirmed nontuberculous mycobacterial lymphadenitis. The specificity of the TST was only 10.5% in these children, with as a consequence false-positive results of NTM infection. In contrast, the specificity for excluding tuberculosis was significantly better using the IGRAs (QFT-G specificity 100%, 95% CI 91–100%, P < 0.001; T-SPOT specificity 98%, 95% CI 87–100%, P < 0.001). It was shown in this analysis that the IGRAs had a higher specificity and, in contrast with the TST, may be used to confirm positive TST results in children in areas with a high incidence of BCG vaccination or NTM infection.

Our meta-analysis also suffers from a number of limitations. Although sensitivity and specificity are useful in assessing the diagnostic value of a test, we are compromised by the lack of a gold standard of latent tuberculosis. It is possible that some individuals enrolled in specificity assessment in this meta-analysis had latent tuberculosis infection, despite the fact that they had no identified risk factors. Longitudinal studies are needed to determine the incidence of active tuberculosis in participants with positive and negative results. According to our strict inclusion criteria, only the commercial tests, QFT-G, QFT-IT and T-SPOT, were within the scope of this analysis, so the number of studies is insufficient and most of them are small. The heterogeneous nature of the methodology also limited the comparability of the studies, so additional studies are needed to better define their performance in diagnosis of pediatric tuberculosis.

Conclusion

Although the results of our analysis should be interpreted with caution, the results could provide useful information to practising clinicians. In addition, most of the studies were published in the past 3 years, making this analysis up to date and timely.

We hope that research will focus on identifying variables that were associated with positive results for each assay in pediatric tuberculosis, for example age, BCG vaccination, contact history and tuberculosis incidence rate of the enrolled countries.

Authors' contribution

L.S. and J.X. contributed equally to this study.

Acknowledgements

A.S. and L.S. conceived and designed the study. L.S. and J.X. performed searches and selected articles according to the inclusion criteria designed in advance. Q.M., W.F., X.W., Q.Y., W.J., C.S., F.L. and D.S. contributed to materials/analysis tools. All authors read and approved the final manuscript. We thank Hugh Nelson and Jifan Hu for revision of the English text. This study was supported in part by grants from the National Natural Science Foundation of China (Nos. 30872788 and 81071315), Beijing Municipal Science Technology Commission (No. Z09050700940903) and Young Scientists fund of Beijing Health Bureau (No. QN2010-025).

References

View Abstract