Risk Factors for Indeterminate Interferon-Gamma Release Assay for the Diagnosis of Tuberculosis in Children—A Systematic Review and Meta-Analysis

Background: Interferon-gamma release assays (IGRA) are well-established immunodiagnostic tests for tuberculosis (TB) in adults. In children these tests are associated with higher rates of false-negative and indeterminate results. Age is presumed to be one factor influencing cytokine release and therefore test performance. The aim of this study was to systematically review factors associated with indeterminate IGRA results in pediatric patients. Methods: Systematic literature review guided by the preferred reporting items for systematic reviews and meta-analyses (PRISMA) searching PubMed, EMBASE, and Web of Science. Studies reporting results of at least one commercially available IGRA (QuantiFERON-TB, T-SPOT.TB) in pediatric patient groups were included. Random effects meta-analysis was used to assess proportions of indeterminate IGRA results. Heterogeneity was assessed using the I2 value. Risk differences were calculated for studies comparing QuantiFERON-TB and T-SPOT.TB in the same study. Meta-regression was used to further explore the influence of study level variables on heterogeneity. Results: Of 1,293 articles screened, 133 studies were included in the final analysis. These assessed QuantiFERON-TB only in 77.4% (103/133), QuantiFERON-TB and T-SPOT.TB in 15.8% (21/133), and T-SPOT.TB only in 6.8% (9/133) resulting in 155 datasets including 107,418 participants. Overall 4% of IGRA results were indeterminate, and T-SPOT.TB (0.03, 95% CI 0.02–0.05) and QuantiFERON-TB assays (0.05, 95% CI 0.04–0.06) showed similar proportions of indeterminate results; pooled risk difference was−0.01 (95% CI −0.03 to 0.00). Significant differences with lower proportions of indeterminate assays with T-SPOT.TB compared to QuantiFERON-TB were only seen in subgroup analyses of studies performed in Africa and in non-HIV-infected immunocompromised patients. Meta-regression confirmed lower proportions of indeterminate results for T-SPOT.TB compared to QuantiFERON-TB only among studies that reported results from non-HIV-infected immunocompromised patients (p < 0.001). Conclusion: On average indeterminate IGRA results occur in 1 in 25 tests performed. Overall, there was no difference in the proportion of indeterminate results between both commercial assays. However, our findings suggest that in patients in Africa and/or patients with immunocompromising conditions other than HIV infection the T-SPOT.TB assay appears to produce fewer indeterminate results.

Background: Interferon-gamma release assays (IGRA) are well-established immunodiagnostic tests for tuberculosis (TB) in adults. In children these tests are associated with higher rates of false-negative and indeterminate results. Age is presumed to be one factor influencing cytokine release and therefore test performance. The aim of this study was to systematically review factors associated with indeterminate IGRA results in pediatric patients.
Methods: Systematic literature review guided by the preferred reporting items for systematic reviews and meta-analyses (PRISMA) searching PubMed, EMBASE, and Web of Science. Studies reporting results of at least one commercially available IGRA (QuantiFERON-TB, T-SPOT.TB) in pediatric patient groups were included. Random effects meta-analysis was used to assess proportions of indeterminate IGRA results. Heterogeneity was assessed using the I 2 value. Risk differences were calculated for studies comparing QuantiFERON-TB and T-SPOT.TB in the same study. Meta-regression was used to further explore the influence of study level variables on heterogeneity.  (1). However, the majority of individuals infected with Mycobacterium tuberculosis are asymptomatic and remain in a latent stage of infection. Data on infected individuals is not included in the World Health Organization TB report as TB infection is not a notifiable disease. Therefore, only estimates exist with one of the most recent estimates suggesting that in 2014 a total of 1.7 billion individuals, equivalent to 23% of the global population, had latent TB infection (2).
Progression from latent TB infection to active TB disease occurs in approximately one in ten adults. Children, however, progress more frequently to active TB and progression may be particularly rapid in the first 2 years of life (3)(4)(5). Early diagnosis and treatment are therefore key to reduce the burden of active TB in children.
Immuno-diagnostic tests are the main tools for the diagnosis of latent TB infection and both the tuberculin skin test (TST) and interferon-gamma release assays (IGRA) are used in the clinical setting (6,7). The latter have been developed to overcome the limited specificity of the TST (8,9). In adults the two commercially available IGRA, the QuantiFERON-TB and T-SPOT.TB-both existing in several test generations-have replaced the TST in many settings, primarily in an attempt to improve specificity (10).
In children, there is evidence that IGRA may have limited sensitivity and therefore the TST is still advocated by most experts (11)(12)(13)(14). In addition, indeterminate IGRA results-due to either high interferon-γ background concentration in the negative control or low interferon-γ response in the positive controlhave been shown to be more frequent in children compared to adults (15)(16)(17)(18).
Underlying reasons for higher proportions of indeterminate IGRA results in children are largely speculative, but several contributing factors including age, concomitant infections and malnutrition have been postulated (18)(19)(20).
The aim of this study was to summarize the existing data on indeterminate IGRA results in children and determine key influencing variables.

Study Selection
A systematic literature search of studies reporting IGRA results in children was performed using PubMed, Embase, and Web of Science. Studies published until October 1st, 2018 were considered. The study was done according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement (21) (Supplementary Material 1 PRISMA Checklist). The following search terms were used: (tuberculosis OR TB) AND [(((t-spot.tb) OR t-spot) OR quantiferon-tb) OR quantiferon] AND (children OR pediatric OR pediatric). The following inclusion criteria were used (i) patients in the pediatric age range with a mean or median age <18 years and a maximum upper age range of 24 years, (ii) results of at least one of the commercially available IGRA detailed (including a statement about indeterminate test results), (iii) publication in English, French, or German. Case reports, case series, conference abstracts and studies involving fewer than 10 participants, commentaries and reviews were excluded. The search and selection of included studies was done by MG, NM, and NR. In unclear cases a joint decision for inclusion or exclusion of the study was made.

Data Extraction
Data were extracted using a standard form including the following variables: year of publication, country in which the study was done, number of participants, mean or median age of participants, age range of participants, type of test performed, number of positive, negative and indeterminate results, definition of indeterminate result, Bacillus Calmette-Guérin (BCG) vaccination status, human immunodeficiency virus (HIV) infection status and information on other potential immunocompromising conditions (e.g., rheumatic diseases, cancer) and concomitant infections (e.g., helminth or other parasitic infections).

Statistical Analysis
The primary outcome was the proportion of indeterminate IGRA results, which was calculated as the number of indeterminate test results divided by the total number of valid test results. Stratified meta-analyses for proportions were performed using a random effects model and the DerSimonian and Laird method, with the estimate of heterogeneity taken from the inverse-variance fixed-effect model. Stratification variables comprised type of IGRA used (QuantiFERON-TB and T-SPOT.TB), age groups (0-7 ≥ 8 years), geographical location of the population under study (Africa, Australia, North America, South America, Asia, Europe) and immune status (HIV infection rate groups, and presence of other immunocompromising factors). Heterogeneity was determined using the I 2 statistic.
In studies comparing both QuantiFERON-TB and T-SPOT.TB, additional stratified meta-analyses for risk differences were performed. Risk differences were defined as the difference in the proportion of indeterminate results between the two IGRA tests and were calculated according to Newcombe and Altman (22). For comparison of pooled risk difference, we applied the DerSimonian and Laird risk difference method. Study weight was indicated by using random effect models for the individual studies to account for the different study characteristics. For risk difference analysis stratification for age groups were done in two groups (0-7 ≥ 8 years) because of the limited number of available datasets.
To further explore potential sources of heterogeneity, we used meta-regression if I 2 was higher than 30%. We considered the following variables as potentially explanatory in a multivariable model: type of IGRA used, age group, geographic location of the population under study, immune status (HIV or other immunocompromising conditions) were considered as explanatory variables in a multivariable model.
We used GraphPad Prism Version 7.02 (GraphPad Software, San Diego, CA, USA) and Stata Version 15.1 (StataCorp, College Station, TX, USA) to generate figures and perform meta-analyses. We reported estimated effect sizes with corresponding 95% confidence intervals (95% CI). A p < 0.05 was considered statistically significant.           Frontiers in Pediatrics | www.frontiersin.org

Demographical Data of the Studies Included
A total of 1,293 citations were identified, of which 379 publications were eligible for full-text assessment and 133 (5 of which were found through additional sources) were included in the final analysis (Figure 1). As 21 publications included data on both QuantiFERON-TB and T-SPOT.TB and one study included data on two different QuantiFERON-TB tests a total of 155 datasets were generated. Table 1  The BCG vaccination rates were reported in 80% (124/155) of datasets and varied from 0 to 100% with a median of 82%. HIV infection rates were reported in 49% (76/155) and varied from 0 to 100% with the median infection rate of 0.05%.

Definition of Indeterminate Results of Interferon-Gamma Release Assays
Definition of indeterminate results for the T-SPOT.TB most commonly referred to low mitogen and/or high nil responses in combination with negative antigen response without stating specific values. Some studies indicated the absolute number of spots as cut-offs, others defined the number of spots in relation to the nil and/or mitogen response. In four studies a nil control of more than 10 spots was considered indeterminate, as opposed to the manufacturer's definition of ≥ 6 spots (92,98,112,145). A total of 21 studies assessed QuantiFERON-TB and T-SPOT.TB in the same study which allowed calculation of risk differences for the proportion of indeterminate results. The pooled proportion of indeterminate results was lower for T-SPOT.TB compared to QuantiFERON-TB (risk difference −0.01, 95% CI −0.03 to −0.00, I 2 = 87.7%), but did not reach statistical significance (Figure 3).

Indeterminate Results According to Age
The mean or median age was specified in 108 datasets; of those 55 datasets had median or mean ages 0-7 years 52 datasets had median or mean ages ≥8years. The pooled proportions of indeterminate results were 0.04 (95% CI 0.03-0.06, I 2 = 94.46%)   For studies which did not specify mean or median age the pooled risk difference was −0.03 (95% CI −0.10 to −0.05, I 2 = 98.5%) (Figure 4). Risk differences within age groups for both assays were not statistically significant.
When continent of the study was included in the risk differences analysis the proportion of indeterminate results for T-SPOT.TB was significantly lower compared to QuantiFERON-TB in studies performed in Africa only (p < 0.001). The pooled risk difference for African studies was −0.022 (95% CI −0.032 to −0.011, I 2 = 15.4%). Risk differences for studies performed on all other continents were not statistically significant (Figures 5, 6).
When immune status was included in the risk difference analysis of indeterminate results the T-SPOT.TB was associated with lower proportions of indeterminate results only in studies that included immunocompromised, HIV-uninfected participants: the pooled risk difference was −0.071(95% CI −0.133 to −0.010, I 2 = 0.0%) and statistically significant (p = 0.022). The risk differences in the remaining groups were not statistically significant (Figure 7).

Meta-Regression of Indeterminate Results
Of the four variables in the model (type of IGRA, age group, continent where study was performed, immune status), only studies including non-HIV-infected immunocompromised patients had a statistically significant contribution to the heterogeneity in the multiple regression model (p = 0.0003).

DISCUSSION
Indeterminate IGRA results have been reported shortly after introduction of these tests in routine clinical use. Despite this, analysis of indeterminate IGRA results has commonly been neglected in the literature, with those results either having been excluded from previous systematic literature reviews or only having been included in very limited subgroup analyses (11,153,154). To our knowledge, this systematic review is the first to comprehensively analyse the occurrence of indeterminate IGRA results in children and adolescents. We found that 4% of IGRA results are indeterminate, suggesting that 1 in 25 tests will not produce a conclusive result. The main factor associated with indeterminate results identified in this meta-analysis was the presence of an immunocompromising condition other than HIV infection.
In our analysis T-SPOT.TB assays were associated with a similar risk of indeterminate results compared to various generations of QuantiFERON-TB tests. T-SPOT.TB assays require lymphocyte adjustment which may reduce the risk of an indeterminate result particularly in patients with reduced lymphocyte count, such as HIV infection or immunocompromising conditions associated with lymphopenia. This assumption is confirmed by results from a meta-analysis including studies in adult HIV-infected patients showing that low CD4 cell counts increased indeterminate rates of QuantiFERON-TB but not of T-SPOT.TB assays (155). Our results contrast with another earlier meta-analysis by Diel et al that reported fewer indeterminate results for QFT-GIT (2.1%) compared to T-SPOT.TB (3.8%) (154). The authors concluded that the more demanding laboratory work for the T-SPOT.TB was likely the reason for higher indeterminate rates. However, their analysis predominately included studies in adults, did not include random effects models, and only included studies published until 2009.
Immunocompromising conditions, including HIV infection, have been identified in earlier studies as a major contributing factor to indeterminate results (16). A study by Oni et al. showed that HIV infection in adults increased the risk of indeterminate results, either through low positive control responses or high interferon-γ background concentrations in the negative control (156). In another study by Mandalakas et al. indeterminate results were more frequent in children infected with HIV than in HIV-uninfected children (116). The previously reported lower sensitivity of QuantiFERON-TB assays in HIV-infected individuals may be linked to a higher rate of indeterminate results, as the difference between the assays was negligible in a study after exclusion of indeterminate results in one analysis (157). Diel et al. reported in their meta-analysis that the rates of indeterminate results for QuantiFERON-TB and T-SPOT.TB assays were higher in immunocompromised compared to immunocompetent individuals, with 6.1 and 4.4%, respectively (154).
Further factors have been shown to influence IGRA results, particularly chronic rheumatic or auto-inflammatory diseases (158,159). IGRA performance depends on intact cellular Th1 responses. Helminth infections, which primarily induce Th2 responses, may alter cytokine production and thereby increase the rate of indeterminate results (20,150,160,161).
Importantly, in our analysis younger age was not associated with indeterminate results, reflected in similar proportions of indeterminate IGRA results in all age groups. This conflicts with several studies that have reported a clear correlation between IGRA performance, proportions of indeterminate results and age (15,16,18,27,33,158). It is well-established that young children have a maturing immune system that may result in diminished cytokine release (162,163). The link between age and cytokine concentrations has also been shown in numerous studies in healthy children unrelated to TB diagnostics (162,163). One potential reason for not detecting a significant association between age and indeterminate IGRA results in this meta-analysis is that aggregate data based on the reported mean/median ages rather than individual patient data were used for this analysis.
There were several factors we were unable to analyse in the datasets that have been reported in some of the included studies, which mainly concern pre-analytical factors. Several studies in children and in adults found a decrease in interferonγ production and indeterminate IGRA results to be associated with delayed sample incubation, shipping of samples, variation in environmental temperatures, and poor phlebotomy technique (164)(165)(166)(167)(168). In addition, co-medication may influence results as a recent ex vivo study showed that both corticosteroids and anti-TNF-alpha agents can cause false-negative IGRA results, and potentially also increase the rate of indeterminate results (169).
One potential limitation of our meta-analysis is the considerable heterogeneity of the included studies. Despite using empirical random effects weighting, excluding studies with < 10 participants, and using only data of the two commercially available IGRAs, heterogeneity remained. Moreover, it is possible that studies with poor IGRA performance and higher proportion of indeterminate results were less likely to be published, leading to publication bias. In addition, details on the type of QuantiFERON-TB assays used were often not reported in the publications, precluding a comparison of different test generations.

CONCLUSIONS
In children, indeterminate IGRA results occur in 1 in 25 tests performed on average. Overall, there was no difference in the proportion of indeterminate results between both commercial assays. However, the data of this meta-analysis indicate that in patients in Africa and/or children with immunocompromising conditions other than HIV infection the T-SPOT.TB assay appears to produce fewer indeterminate results than the QuantiFERON-TB assays.

AUTHOR CONTRIBUTIONS
NR, MT, and UH conceptualized the study. NM and MG designed the search strategy and searched the literature, selected the studies and extracted the data. NR reviewed and approved the search strategy. NM, MG, and TV performed the data analysis. All authors performed the data interpretation. NM, MG, and NR wrote the draft manuscript. All authors reviewed, provided intellectual input into and approved the final manuscript.