BACKGROUND: Interferon-γ release assays (IGRAs) are important adjunctive tests for diagnosing tuberculosis (TB) disease in children.
METHODS: We analyzed California TB registry data for patients ≤18 years with laboratory-confirmed TB disease during 2010–2015 to identify case characteristics associated with test selection and performance and measure IGRA sensitivity.
RESULTS: In total, 778 cases of TB were reported; 360 were laboratory confirmed. Indeterminate IGRAs were associated with being <1 year old (prevalence rate ratio 9.23; 95% confidence interval 2.87 to 29.8) and having central nervous system disease (prevalence rate ratio 2.69; 95% confidence interval 1.06 to 6.86) on multivariable analysis. Ninety-five children had both an IGRA and tuberculin skin test (TST) performed. Among those, the sensitivity of IGRA in 5- to 18-year-olds was 96% (66 out of 69) vs 83% (57 out of 69) for TST (P = .01); IGRA sensitivity compared with TST in children ages 2 to 4 was 91% (10 out of 11) vs 91% (10 out of 11) (P > .99), and the sensitivity compared with TST in children aged <2 years was 80% (12 out of 15) vs 87% (13 out of 15) (P > .99).
CONCLUSIONS: This is the largest North American analysis of IGRA use and performance among children with TB disease. In children <5 years old, IGRA sensitivity is similar to TST, but sensitivity of both tests are reduced in children <2 years old. Indeterminate results are higher in children <1 year old and in central nervous system disease. In children ≥5 years old with laboratory-confirmed TB, IGRA has greater sensitivity than TST and should be considered the preferred immunodiagnostic test..
- CDC —
- Centers for Disease Control and Prevention
- CDPH —
- California Department of Public Health
- CI —
- confidence interval
- CNS —
- central nervous system
- IGRA —
- interferon-γ release assay
- PRR —
- prevalence rate ratio
- QFT —
- TB —
- TST —
- tuberculin skin test
What’s Known on This Subject:
Interferon-γ release assays (IGRAs) and tuberculin skin tests (TSTs) are important adjunctive pediatric tuberculosis (TB) diagnostics. Data on IGRA and TST sensitivity in laboratory-confirmed TB disease has been hampered by the rarity of laboratory confirmation in pediatric TB.
What This Study Adds:
This population-based analysis reveals superior sensitivity of IGRAs versus TSTs in children ≥5 years old with confirmed TB and similar sensitivity in children <5 years old. Children <1 year or with central nervous system disease were more likely to have indeterminate IGRAs.
Because microbiologic diagnostics are inadequately sensitive, the diagnosis of tuberculosis (TB) in children is often based on clinical symptoms, epidemiologic risk, chest radiographs, and immunologic tests of infection.1,2 In low-incidence countries, an important factor to consider is known exposure to a TB case, but children often present as the index case.3,4 As such, indirect immunologic tests like the tuberculin skin test (TST) and interferon-γ release assay (IGRA) are used to contribute important information to the decision to initiate therapy for pediatric TB disease.5
Aside from microbiologic confirmation, there is no reference standard for TB disease. Historically, the TST has been used to support the diagnosis in cases in which TB is suspected. Nevertheless, in severe forms of TB, TST sensitivity is limited; for example, only 60% of patients with proven TB meningitis receive TST positive results.6 TST specificity is also suboptimal, because nontuberculous mycobacterial infection and BCG vaccination can cause false-positive TST results.7 IGRAs have superior test specificity to TST, but data on their sensitivity in children with TB disease, especially those <5 years of age, remains limited. Interpretation of current literature is challenging because laboratory-confirmed cases are uncommon and definitions for clinical disease are not standardized.8–14 The authors of 3 meta-analyses have found comparable sensitivity of IGRA and TST in children with TB disease.15–17
In this retrospective, population-based analysis, we analyzed TST and IGRA results for all cases of pediatric TB disease reported in California during 2010–2015. The primary goals were to determine existing patterns of test usage, evaluate the impact of patient age and disease site on test performance, and assess test sensitivity in children with laboratory-confirmed disease.
Cases of TB disease in patients ≤18 years of age during 2010–2015 were analyzed. Reporting TB is mandated by law (California Code of Regulations Title 17-2500). TB cases are reported to the California Department of Public Health (CDPH) TB registry by local health jurisdictions according to case definitions set by the Centers for Disease Control and Prevention (CDC).18 The TB registry is quality controlled and includes data about demographic and clinical characteristics, diagnostic test results, treatment, and outcomes. BCG history was assigned on the basis of country of birth by referencing the BCG World Atlas.19 This analysis was conducted in accordance with the CDPH mandate to evaluate surveillance data for public health purposes and did not require human subject review.
A positive TST result was defined as having ≥5 mm of induration. Site of disease was defined into mutually exclusive categories as pulmonary only, any central nervous system (CNS), non-CNS extrapulmonary, and multifocal, non-CNS disease. Laboratory-confirmed disease included children with a result positive for acid-fast bacilli smear, nucleic acid amplification test, or culture for TB. All cases lacking confirmation were categorized as clinical.
TB cases evaluated with a TST, IGRA, or both were compared by using a χ2 test (Table 1). Bivariate prevalence rate ratios (PRRs) and 95% confidence intervals (CIs) were calculated with Cochran–Mantel–Haenszel logit estimates (Table 2). Multivariable PRRs and 95% CIs were calculated with a generalized linear model by using a logit link function and Poisson distribution with robust SEs.20 We constructed the multivariable model including variables based on biologic plausibility (age, immunosuppression) as well as those with at least 1 strata with a bivariate 95% CI that did not cross 1. IGRA and TST concordance was assessed among children that had both tests by percentage agreement and statistical concordance by using an unweighted Cohen’s κ score excluding indeterminate IGRAs (Table 3). Age and site of disease were assessed for association with indeterminate IGRAs by using bivariate χ2 or Fisher’s exact tests (Table 4). Sensitivity was calculated for IGRAs and TSTs and compared by using a McNemar’s exact test among those who received both TST and IGRA, and Z scores were used for comparison among all cases (Table 5). Indeterminate results were handled in the following 2 ways in sensitivity analyses: we excluded cases with indeterminate results and categorized cases with negative results. Statistical analysis was performed with SAS 9.3 software (SAS Institute, Inc, Cary, NC). Z scores were calculated by using Stata (Stata Corp, College Station, TX).
Test Use by Disease Characteristics and Demographics
A total of 778 cases of TB in children with TST or IGRA results were reported in California during 2010–2015 (Table 1, Fig 1). Of these, 456 (58.6%) had a TST, 117 (15%) had an IGRA, and 205 (26.4%) had both. IGRAs included 318 QuantiFERON (QFT) (enzyme-linked immunosorbent assay-based) tests and 4 T-Spot.TB (enzyme-linked immunospot-based) tests. One child had both IGRA types done with concordant results. Children with laboratory-confirmed disease (n = 360) were less likely to have a TST (192 out of 456; 42%) than an IGRA (69 out of 117; 59%) or both (99 out of 205 [48%]; P = .004). Being <2 years of age (n = 144) and being 2 to 4 years of age (n = 171) was associated with use of TST (108 out of 456 [24%] and 113 out of 456 [25%], respectively), and when an IGRA was performed, it was more often obtained with a TST (P < .001) compared with children ≥5 years old. The median age was 15 years for patients who received an IGRA only, 5 years for patients who received a TST only, and 9 years for patients who received both tests.
Of 208 non–US-born children, 207 were born in countries that give routine BCG vaccinations. Non–US-born children were more likely to receive an IGRA or both tests (46 out of 117 [39%] and 59 out of 205 [29%], respectively) than a TST only (103 out of 456 [23%]; P < .001). Children tested as part of a contact investigation were more likely to receive only a TST (153 out of 456; 34%) versus an IGRA (17 out of 117; 15%) and both (21 out of 205 [10%]; P < .001), than those presenting with TB symptoms or an abnormal chest radiograph, who were more likely to receive an IGRA (82 out of 117; 70%) or both tests (152 out of 205; 74%) versus TST (262 out of 456 [57.5%]; P < .001).
Predictors of Positive IGRA Results
A positive TST result was associated with a positive IGRA result in laboratory-confirmed cases (PRR 1.56, 95% CI 1.07 to 2.26; Table 2). Size of the TST beyond 5 mm of induration was not associated with a positive IGRA result. Being <2 years of age was associated with a decreased likelihood of a positive IGRA result in bivariate (PRR 0.69; 95% CI 0.52 to 0.92) and multivariable (PRR 0.72; 95% CI 0.55 to 0.93) comparisons with patients aged 5 to 18 years. Of children aged 2 to 4 years, 13 out 16 (81%) had a positive IGRA result compared with 119 out of 126 (94%) of patients aged 5 to 18 years, but this difference was nonsignificant. All 23 patients with multifocal, non-CNS disease had a positive IGRA result (PRR 1.12; 95% CI 1.05 to 1.21). Only 11 out of 17 (65%) children with CNS disease had a positive IGRA result, but this was not significantly different from pulmonary-only disease.
IGRA and TST Concordance
In laboratory-confirmed TB there was a greater concordance between TST and IGRA in children <2 years of age compared with patients 2 to 4 and 5 to 18 years of age (14 out of 15 [93%] vs 9 out of 11 [82%] and 58 out of 69 [84%], respectively; P ≤ .001) (Table 3). No difference was observed related to country of origin or site of disease. There was an overall high level of concordance between the TST and IGRA among children receiving both tests. Among clinically diagnosed cases, there was nonsignificantly lower agreement in non–US-born patients (20 out of 26 [77%] vs 66 out of 76 [87%]; P = .06) and lower concordance in patients ˂2 years and 2 to 4 years versus 5 to 18 years (6 out of 8 [75%] vs 22 out of 29 [76%] and 58 out of 65 [89%], respectively; P = .05) (Supplemental Table 6).
Patient and Disease Characteristics Associated With Indeterminate IGRA Results
There were 12 reported cases with indeterminate IGRA results, 8 of which had a TST performed (Table 4). When compared with children who received positive or negative results, children with indeterminate IGRA were more likely to receive a negative TST result (PRR 5.03; 95% CI 1.32 to 19.1). Young age (<1 year old) and CNS disease were associated with an increased likelihood of indeterminate results in both bivariate and multivariable analysis. The relationship with age was stepwise: 3 out of 13 (23%) children <1 year of age, 2 out of 23 (9%) children aged 1 year, and 3 out of 58 (5%) children 2 to 4 years had indeterminate results compared with 4 out of 228 (2%) patients aged 5 to 18 years. In children ˂1 year old, 3 out of 4 (75%) with CNS disease had indeterminate IGRA results compared with 0 out of 9 infants with other disease sites. No indeterminate IGRA results (0 of 18) occurred in children with CNS disease who were ≥2 years old (Supplemental Table 7).
Sensitivity of IGRA and TST in Laboratory-Confirmed TB Disease
Sensitivity of IGRA and TST were compared in children with laboratory-confirmed disease (Table 5). Among children with both tests performed (n = 95; 4 indeterminate IGRA results excluded) there was no significant difference in sensitivity for TST compared with IGRA (80 out of 95 [84%] vs 88 out of 95 [93%]; P = .6). Using results from both tests provided an increase in test sensitivity compared with TST alone (91 out of 95 [96%] vs 80 out of 95 [84%]; P < .001) but not compared with IGRA alone (91 out of 95 [96%] vs 88 out of 95 [93%]; P = .25). There was no difference in sensitivity when stratified by non-US birth, and performance of both TST and IGRA offered increased sensitivity compared with TST alone regardless of US or non-US birth. IGRA sensitivity compared with TST in children 2 to 4 was 91% (10 out of 11) vs 91% (10 out of 11) with P > .99 and 80% (12 out of 15) vs 87% (13 out of 15) with P > .99 in children <2 years of age. Performing both tests did not improve overall sensitivity in these age groups. In patients 5 to 18 years of age, TST had inferior test sensitivity in laboratory-confirmed disease compared with IGRA (57 out of 69 [83%] vs 66 out of 69 [96%]; P = .01). In this age range, performance of TST and IGRA improved sensitivity compared with TST (67 out of 69 [97%] vs 57 out of 69 [83%]; P = .002) but not compared with IGRA alone (67 out of 69 [97%] vs 66 out of 69 [96%]; P > .99). Test sensitivity did not differ significantly when stratified by site of disease. Additional analyses of test sensitivity were performed (Supplemental Table 8).
Among all cases with ≥1 test performed, the sensitivity of the IGRA (149 out of 169; 93%) was higher than for TST (254 out of 291 [87.3%]; P = .05). IGRA had higher sensitivity than TST among US-born patients (99 out of 104 [95%] vs 183 out of 209 [88%]; P = .03) and patients 5 to 18 years old (119 out of 125 [95%] vs 176 out of 198 [89%]; P = .05). Among non-US born children and children <5 years old, there was no difference in test sensitivity. IGRA test sensitivity was comparable in children 2 to 4 years versus patients 5 to 18 years (13 out of 14 [93%] vs 119 out of 125 [95%]) but both TST and IGRA had decreased sensitivity in children <2 years (43 out of 55 [78%] and 17/21 [81%], respectively).
With this population-based analysis, we describe IGRA use and sensitivity in laboratory-confirmed pediatric TB. In this cohort, 46% of children had laboratory-confirmed TB vs 34% of culture-confirmed cases in a US population-based analysis of pediatric TB (ages 0–18 years).21 This high rate of laboratory-confirmed disease makes this an ideal population to study IGRA performance because our laboratory-confirmed cases capture the largest possible spectrum of pediatric TB disease.
Our findings indicate that, consistent with American Academy of Pediatrics recommendations, children ˂5 years of age were less likely to be evaluated by IGRA. Also consistent with American Academy of Pediatrics and CDC guidelines, non-US birth was associated with an increased use of IGRA and reflects clinician awareness of the superior specificity of IGRA for BCG-vaccinated persons.5,22–25 However, 22% of non–US-born, BCG-vaccinated children were tested with TST alone suggesting the IGRA may still be underused in this population. Children presenting with TB-related symptoms were more likely to be tested with IGRA, whereas children evaluated as part of a contact investigation were more often tested with a TST. These data, as well as differences in disease presentation, might reflect test availability and use in public health versus hospital settings. Additionally, our finding of higher IGRA test sensitivity versus TST in 5- to 18-year-old children, coupled with known superior specificity, may indicate the need to increase access to IGRA testing.
Compared with children with disease restricted to a single site, children with multifocal, non-CNS TB were more likely to receive IGRA positive results. Conversely, children with CNS disease had an increased likelihood of both negative TST and IGRA results. Children with CNS disease have lower rates of TST positivity than in children with non-CNS disease,6 and we found similar results with IGRA. CNS disease was also associated with a higher rate of indeterminate IGRA results independent of age, suggesting that, in a child with CNS symptoms, an indeterminate IGRA result must not be considered evidence against a diagnosis of TB. Likewise, authors of a recent analysis (of adults and children) found greater odds of disseminated disease and death with an indeterminate IGRA result.26
One of the primary uncertainties surrounding the IGRA is test sensitivity in children <5 and particularly in children <2 years of age. In this analysis, we evaluated the impact of age on IGRA results in multiple ways. First, being <2 years of age was associated with a decreased likelihood of a positive IGRA result in laboratory-confirmed disease compared with older children after controlling for site of disease. However, IGRA in children <2 years of age was not significantly less sensitive than TST (12 out 15 [80%] vs 13 out of 15 [87%]); both tests performed less well in this age range. Among children aged 2 to 4 years, IGRA performed similarly to TST, and both tests had sensitivity of ˃90% in laboratory-confirmed TB.
Authors of previous reports describe widely variable rates of QFT indeterminate results in relation to age, with some previous analyses revealing a clear association with young age.27–29 In a more recent analysis of infants <2 years old, authors reported an indeterminate rate of 4.2% without an age association,30 and others have reported indeterminate results of <1% in children.31,32 In a large trial of ˃2500 infants, authors reported an overall indeterminate rate of 0.4%.33 Factors that might influence indeterminate results include host immune status and comorbidities, specific batch and generation of IGRA used, and preanalytic processing of the blood specimen.34 Our analysis revealed a clinically relevant association between indeterminate IGRA results and young age in a real-world context. Nevertheless, although we adjusted for CNS disease, the strong association of CNS disease and young age likely contributed to the increased risk of an indeterminate result in young children because of residual confounding. In fact, 4 of 5 children <2 years old with indeterminate IGRA results had CNS disease suggesting that CNS disease may be a stronger driver of indeterminate test results than age.
The concordance between IGRA and TST in children <2 years of age with laboratory-confirmed disease was excellent at 93% (κ score = 0.76). In fact, concordance was stronger in this age group than in any other, suggesting that IGRAs perform comparably to TSTs. Consistent with several recent meta-analyses,15–17 we found IGRA sensitivity in children <2 years is equivalent to TST; in other words, a false-negative IGRA result is no more likely than a false-negative TST result. We found that discordant results of IGRA−/TST+ was higher in clinically diagnosed TB, which may reflect a positive TST result influencing the clinical diagnosis of TB.7
Although performing both tests improved test sensitivity over TST overall and in certain subgroups (US-born and 5–18-year-olds), there was no gain in sensitivity when performing both tests versus IGRA overall or in any subgroup. This finding differs from previous studies9,35,36 which included both laboratory-confirmed and clinical diagnoses and may have been influenced by the impaired specificity of TST. However, we agree with the approach recommended by guidelines of obtaining both tests simultaneously when optimal sensitivity is desired.5,23 In such situations, when evaluating for TB disease, we suggest with our data that an IGRA is likely to add more as an additional diagnostic than a TST.
IGRA performed with greater sensitivity in children ≥5 years of age with laboratory-confirmed disease. IGRA alone identified 96% of cases, whereas TST alone identified 83% of cases. These data, which are in concert with the known superiority in test specificity,7,24,37 suggest that IGRA should be the preferred test in children 5 to 18 years of age who are being evaluated for TB disease. In this cohort, only 49.2% (228 out of 463) of 5- to 18-year-old children and 50% (105 out of 208) of non–US-born children received an IGRA indicating the test is underused. The superior specificity of IGRA in BCG-vaccinated children is well known, butthese data suggestthat IGRA may havesuperior sensitivity to TST. This provides further rationale for making IGRA the test of choice in all 5- to 18-year-old patients being evaluated for TB disease.
Our analysis has several limitations. Our cohort was tested predominately with enzyme-linked immunosorbent assay-based IGRA (QFT) and so provides little insight into the performance of enzyme-linked immunospot tests (T-SPOT.TB). The generalizability of the sensitivity data is limited by differences in patient characteristics among children receiving both IGRA and TST versus children receiving a single test. We did not have access to quantitative IGRA results and were not able to investigate alternate IGRA cut points or reasons for indeterminate results. Additionally, we were unable to record multiple IGRA results, so a patient with an indeterminate result followed by a valid result would not be captured. We did not consider timing of TST in relation to IGRA; however, the influence of TST on subsequent IGRA is likely to be minimal in patients with active TB. Although this was the largest North American population-based analysis of IGRA performance in children of all ages and children <5 years of age with laboratory-confirmed TB, the cohort size of children <5 years of age (n = 42) remains a limitation.
IGRA had higher sensitivity compared with TST in Californian children 5 to 18 years of age with laboratory-confirmed TB disease and similar sensitivity among children <5 years of age. Given the tests’ advantages in specificity as well, we argue that an IGRA should be considered the test of choice when evaluating children 5 to 18 years old for TB disease in high-resource, low-TB burden settings regardless of BCG status. Our data reveals that there is currently underuse of IGRAs in children 5 to 18 years old. In 2- to 4-year-old children, the IGRA and TST had similar sensitivities of >90%, whereas, in children ˂2 years of age, both tests had reduced sensitivity of ∼80%. Because of the overall poor sensitivity of both tests and increased proportion of indeterminate IGRA results, coupled with severe consequences of missing TB in a sick child ˂2 years old, negative or indeterminate test results should not delay empirical TB treatment when otherwise indicated.
This work fell under the mandate of the CDPH to collect and analyze surveillance data on TB in California. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the CDC or the Department of Health and Human Services.
- Accepted March 6, 2018.
- Address correspondence to Pennan M. Barry, MD, MPH, California Department of Public Health, Tuberculosis Control Branch, 2nd Floor, Building P, 850 Marina Bay Pkwy, Richmond, CA 94804. E-mail:
FINANCIAL DISCLOSURE: Dr Islam has received speaker honoraria and travel support from the Qiagen (makers of QuantiFERON); the other authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by the grant or cooperative agreement U52PS004656; funded by the Centers for Disease Control and Prevention.
POTENTIAL CONFLICT OF INTEREST: Dr Islam has received speaker honoraria and travel support from the Qiagen (makers of QuantiFERON); the other authors have indicated they have no potential conflicts of interest to disclose.
- Kampmann B,
- Whittaker E,
- Williams A, et al
- Bamford ARJ,
- Crook AM,
- Clark JE, et al
- Centers for Disease Control and Prevention
- U.S. Department of Health and Human Services
- Fang J
- Mazurek GH,
- Jereb J,
- Vernon A,
- LoBue P,
- Goldberg S,
- Castro K; IGRA Expert Committee; Centers for Disease Control and Prevention
- Lewinsohn DM,
- Leonard MK,
- LoBue PA, et al
- Auld SC,
- Lee SH,
- Click ES, et al
- Haustein T,
- Ridout DA,
- Hartley JC, et al
- Garazzino S,
- Galli L,
- Chiappini E, et al; SITIP IGRA Study Group
- Howley MM,
- Painter JA,
- Katz DJ, et al; Tuberculosis Epidemiologic Studies Consortium
- Andrews JR,
- Nemes E,
- Tameris M, et al
- Hill PC,
- Brookes RH,
- Adetifa IM, et al
- Lighter J,
- Rigaud M,
- Eduardo R,
- Peng CH,
- Pollack H
- Copyright © 2018 by the American Academy of Pediatrics