OBJECTIVE: To evaluate the performance of QuantiFERON-TB GOLD (QFTG) in a resource-poor setting among patients with and without HIV infection. DESIGN: Cross-sectional study. SETTING: Two hospitals in Northern Tanzania. SUBJECTS: Eighty three adult male and female inpatients. INTERVENTION: All patients were screened for HIV infection and underwent tuberculin skin test (TST) and QFTG. RESULTS: Eighty-three subjects were enrolled, and 29 (35%) of 83 were HIV-infected. QFTG yielded indeterminate results in 12 (22%; 95% CI 12%-34%) of 54 HIV-uninfected and 13 (45%; 95% CI 26%-64%) of 29 HIV-infected subjects (p = 0.0323). Among those with smear-positive pulmonary tuberculosis, TST was positive in 40 (100%; 95% CI 91%-100%) of 40 HIV-uninfected subjects compared with seven (54%; 95% CI 25%-81%) of 13 HIV-infected subjects (p < 0.0001), and QFTG was positive in 28 (70%; 95% CI 53%-83%) of 40 HIV-uninfected subjects compared with three (23%; 95% CI 5%-54%) of 13 HIV-infected subjects (p = 0.0029). Among medical inpatients at risk for latent tuberculosis infection, TST was positive in seven (50%) of 14 HIV-uninfected patients and three (19%) of 16 HIV-infected patients (p = 0.0701) and QFTG was positive among two (14%) of 14 HIV-uninfected patients and three (19%) of 16 HIV-infected patients (p = 0.7437). CONCLUSIONS: The presence of HIV co-infection was associated with a significant reduction in sensitivity of both the TST (p < 0.0001) and QFTG (p = 0.0029) for the diagnosis of active M. tuberculosis infection. The high proportion of indeterminate QFTG and lack of sensitivity, particularly among HIV-infected patients, may limit its applicability in settings like Tanzania. Larger studies in resource-poor settings are required.