Abstract
RATIONALE: Two forms of the IFN-gamma release assay (IFNGRA) to detect tuberculosis infection are available, but neither has been evaluated in comparable HIV-infected and uninfected persons in a high tuberculosis incidence environment. OBJECTIVE: To compare the ability of the T-SPOT.TB (Oxford Immunotec, Abingdon, UK), QuantiFERON-TB Gold (Cellestis, Melbourne, Australia), and Mantoux tests to identify latent tuberculosis in HIV-infected and uninfected persons. METHODS: A cross-sectional study of 160 healthy adults without active tuberculosis attending a voluntary counseling and testing center for HIV infection in Khayelitsha, a deprived urban South African community with an HIV antenatal seroprevalence of 33% and a tuberculosis incidence of 1,612 per 100,000. MEASUREMENTS AND MAIN RESULTS: One hundred and sixty (74 HIV(+) and 86 HIV(-)) persons were enrolled. A lower proportion of Mantoux results was positive in HIV-infected subjects compared with HIV-uninfected subjects (p < 0.01). By contrast, the proportion of positive IFNGRAs was not significantly different in HIV-infected persons for the T-SPOT.TB test (52 vs. 59%; p = 0.41) or the QuantiFERON-TB Gold test (43 and 46%; p = 0.89). Fair agreement between the Mantoux test (5- and 10-mm cutoffs) and the IFNGRA was seen in HIV-infected people (kappa = 0.52-0.6). By contrast, poor agreement between the Mantoux and QuantiFERON-TB Gold tests was observed in the HIV-uninfected group (kappa = 0.07-0.30, depending on the Mantoux cutoff). The pattern was similar for T-SPOT.TB (kappa = 0.18-0.24). Interpretation: IFNGRA sensitivity appears relatively unimpaired by moderately advanced HIV infection. However, agreement between the tests and with the Mantoux test varied from poor to fair. This highlights the need for prospective studies to determine which test may predict the subsequent risk of tuberculosis.
