Optimizing the detection of recent tuberculosis infection in children in a high tuberculosis-HIV burden setting

Abstract

Rationale: Children who are young, malnourished, and infected with HIV have significant risk of tuberculosis (TB) morbidity and mortality following TB infection. Treatment of TB infection is hindered by poor detection and limited pediatric data.

Objectives: Identify improved testing to detect pediatric TB infection.

Methods: This was a prospective community-based study assessing use of the tuberculin skin test and IFN-γ release assays among children (n = 1,343; 6 mo to <15 yr) in TB-HIV high-burden settings; associations with child characteristics were measured.

Measurements and main results: Contact tracing detects TB in 8% of child contacts within 3 months of exposure. Among children with no documented contact, tuberculin skin test and QuantiFERON-TB Gold In-Tube positivity was greater than T-SPOT.TB. Nearly 8% of children had IFN-γ release assay positive and skin test negative discordance. In a model accounting for confounders, all tests correlate with TB contact, but IFN-γ release assays correlate better than the tuberculin skin test (P = 0.0011). Indeterminate IFN-γ release assay results were not associated with age. Indeterminate QuantiFERON-TB Gold In-Tube results were more frequent in children infected with HIV (4.7%) than uninfected with HIV (1.9%), whereas T-SPOT.TB indeterminates were rare (0.2%) and not affected by HIV status. Conversion and reversion were not associated with HIV status. Among children infected with HIV, tests correlated less with contact as malnutrition worsened.

Conclusions: Where resources allow, use of IFN-γ release assays should be considered in children who are young, recently exposed, and infected with HIV because they may offer advantages compared with the tuberculin skin test for identifying TB infection, and improve targeted, cost-effective delivery of preventive therapy. Affordable tests of infection could dramatically impact global TB control.

Keywords: HIV; IFN-γ release tests; latent tuberculosis infection; pediatrics; tuberculin test.

Figure 1.

Tuberculin skin test (TST) distribution stratified by HIV status and tuberculosis (TB) exposure. (A) Data from TB-exposed (red bars) and TB-unexposed (black bars) children infected with HIV and two normalized lines for recent contacts (red line) and nonexposed control subjects (black line). (B) Data from TB-exposed (red bars) and TB-unexposed (black bars) children not infected with HIV and two normalized lines for recent contacts (red line) and nonexposed control subjects (black line). The normalized lines are derived using polynomial (Poly.) expressions because this fits our bell-shaped curve data best. Nonresponsive children were excluded from the histogram but varied by HIV and exposure status: HIV-infected and TB-exposed (53%); HIV-infected and no documented TB exposure (70%); HIV-uninfected and TB-exposed (47%); HIV-uninfected and no documented TB exposure (62%).

Figure 2.

Test positivity and treatment implications in relationship to contact score. Bars represent the proportion of children with respective positives test results along the continuum of exposure ranging from no known exposure (contact score = 0) to the highest level of exposure (contact score = 10). Lines demonstrate the cumulative proportion of children that would be offered IPT based on test positivity if contact score were considered a cut-point for IPT eligibility. If testing were used to guide IPT among children with a contact score greater than or equal to 2, TST, Tspot, and QFT would result in 26, 21, and 29% of children being offered treatment, respectively. In contrast, if testing were used to guide IPT among children with a contact score of greater than or equal to 5, TST, Tspot, and QFT would result in 20, 16, and 22% of children being offered treatment. Finally, if testing were used to guide IPT among children with a contact score greater than or equal than 10, no child would be treated because the highest contact score possible is 10. IPT = isoniazid preventive therapy; QFT = QuantiFERON; RX QFT = treatment guided by QFT response; RX Tspot = treatment guided by Tspot response; RX TST = treatment guided by TST response; Tspot = T-Spot.TB; TST = tuberculin skin test.

Figure 3.

Regression slopes for exposure continuum expressed as contact score. To visually represent the association between the contact score and test outcome, the odds ratio for each positive level of the score versus a zero score was estimated from each adjusted logistic regression model. The natural logarithm of the odds ratio was then plotted against the contact score to allow for comparison between the tests. The slopes of the lines are estimated from the regression of the logs of odds ratios of each successive higher exposure compared with the least exposed group. Hence, a steeper slope represents a greater change in the log odds ratio as exposure increases. A greater change in the log odds ratio in response to increasing exposure (i.e., steeper slope) indicates that test results are better correlated with the likelihood of infection. This finding suggests that tests with a steeper slope are better able to detect infection. QFT = QuantiFERON; Tspot = T-Spot.TB; TST = tuberculin skin test.

Figure 4.

Odds ratio of test positivity in young children compared with older children. The figure compares the aOR of test positivity for each test in children younger than 24 months compared with children greater than or equal to 24 months. The aOR are indicated by the circles and the 95% confidence intervals are illustrated with the lines; Tspot in red, TST in black, and QFT in gray. For the QFT model and the Tspot, there were interactions between age and contact score. Hence, the association between the IFN-γ release assay results and contact score is most accurately measured within each of the respective age strata. The aOR of QFT positivity was higher for younger children (P < 0.001): children younger than 24 months, aOR of 1.44 (95% CI, 1.27–1.64) and children greater than or equal to 24 months, aOR of 1.11 (95% CI, 1.06–1.17). Similarly, the aOR of Tspot positivity was higher for younger children (P < 0.001): children younger than 24 months, aOR of 1.50 (95% CI, 1.30–1.74) and children greater than or equal to 24 months, aOR of 1.12 (95% CI, 1.06–1.20). For the TST model, there was no significant association between age and TST result. Therefore, the aOR of TST was 1.16 (95% CI, 1.10, 1.22) for children younger than and greater than or equal to 24 months. aOR = adjusted odds ratio; CI = confidence interval; QFT = QuantiFERON; Tspot = T-Spot.TB; TST = tuberculin skin test.