BACKGROUND: Children with latent tuberculosis infection (LTBI) represent a huge reservoir for future disease. We wished to determine Mycobacterium tuberculosis (M.tb) infection prevalence among BCG-immunised five-year-old children in Entebbe, Uganda, but there are limited data on the performance of immunoassays for diagnosis of tuberculosis infection in children in endemic settings. We therefore evaluated agreement between a commercial interferon gamma release assay (T-SPOT.TB) and the tuberculin skin test (TST; 2 units RT-23 tuberculin; positive defined as diameter ≥10 mm), along with the reproducibility of T-SPOT.TB on short-term follow-up, in this population. METHODOLOGY/PRINCIPAL FINDINGS: We recruited 907 children of which 56 were household contacts of TB patients. They were tested with T-SPOT.TB at age five years and then re-examined with T-SPOT.TB (n = 405) and TST (n = 319) approximately three weeks later. The principal outcome measures were T-SPOT.TB and TST positivity. At five years, 88 (9.7%) children tested positive by T-SPOT.TB. More than half of those that were T-SPOT.TB positive at five years were negative at follow-up, whereas 96% of baseline negatives were consistently negative. We observed somewhat better agreement between initial and follow-up T-SPOT.TB results among household TB contacts (κ = 0.77) than among non-contacts (κ = 0.39). Agreement between T-SPOT.TB and TST was weak (κ = 0.28 and κ = 0.40 for T-SPOT.TB at 5 years and follow-up, respectively). Of 28 children who were positive on both T-SPOT.TB tests, 14 (50%) had a negative TST. Analysis of spot counts showed high levels of instability in responses between baseline and follow-up, indicating variability in circulating numbers of T cells specific for certain M.tb antigens. CONCLUSIONS/SIGNIFICANCE: We found that T-SPOT.TB positives are unstable over a three-week follow-up interval, and that TST compares poorly with T-SPOT.TB, making the categorisation of children as TB-infected or TB-uninfected difficult. Existing tools for the diagnosis of TB infection are unsatisfactory in determining infection among children in this setting.