OBJECTIVES: To determine burden and risk factors for tuberculosis (TB) in an antiretroviral treatment (ART) programme and its impact on ART outcomes. DESIGN: Prospective cohort study. METHODS: Prevalent TB was assessed at baseline and incident TB was ascertained prospectively over 3 years among 944 patients accessing a community-based ART programme in South Africa. RESULTS: At enrollment, median CD4 cell count was 96 cells/microl and 52% of patients had a previous history of TB. Prevalent TB (current antituberculosis treatment or active TB) was present in 25% and was strongly associated with advanced immunodeficiency. During 782 person-years of ART, 81 cases of TB were diagnosed. The incidence was 22.1/100 person-years during the first 3 months of ART and decreased to an average of 4.5/100 person-years during the second and third years. In multivariate analysis, risk of incident TB during follow-up was only associated with the current absolute CD4 cell count at that time point; an increase of 100 cells/mul was associated with a 25% lower risk (P = 0.007). Although prevalent and incident TB were associated with greater than two-fold increased mortality risk, they did not compromise immunological and virological outcomes among survivors at 48 weeks. CONCLUSIONS: Late initiation of ART was associated with a major burden of TB in this ART programme. TB reduced survival but did not impair immunovirological outcomes. Reductions in TB incidence during ART were dependent on CD4 cell count; however, after 3 years of treatment, rates were still 5- to 10-fold higher than among non-HIV-infected people. Earlier initiation of ART may reduce this burden of TB.