Objective: To analyze tuberculosis treatment outcomes and their predictors. Methods: This was a retrospective longitudinal cohort study involving tuberculosis patients treated between 2004 and 2006 at the Instituto de Pesquisa Evandro Chagas, in the city of Rio de Janeiro. We estimated adjusted risk ratios (ARRs) for the predictors of treatment outcomes. Results: Among 311 patients evaluated, the rates of cure, treatment abandonment, treatment failure, and mortality were 72%, 19%, 2%, and 6%, respectively. Changes in the treatment regimen due to adverse events occurred in 8%. The factors found to reduce the probability of cure were alcoholism (ARR, 0.30), use of the streptomycin+ethambutol+ofloxacin (SEO) regimen (ARR, 0.32), HIV infection without the use of antiretroviral therapy (ART; ARR, 0.36), and use of the rifampin+isoniazid+pyrazinamide+ethambutol regimen (ARR, 0.58). Being younger and being alcoholic both increased the probability of abandonment (ARR, 3.84 and 1.76, respectively). It was impossible to determine the ARR for the remaining outcomes due to their low prevalence. However, using the relative risk (RR), we identified the following potential predictors of mortality: use of the SEO regimen (RR, 11.43); HIV infection without ART (RR, 9.64); disseminated tuberculosis (RR, 9.09); lack of bacteriological confirmation (RR, 4.00); diabetes mellitus (RR, 3.94); and homosexual/bisexual behavior (RR, 2.97). Low income was a potential predictor of treatment failure (RR, 11.70), whereas disseminated tuberculosis and HIV infection with ART were potential predictors of changes in the regimen due to adverse events (RR, 3.57 and 2.46, respectively). Conclusions: The SEO regimen should not be used for extended periods. The data confirm the importance of ART and suggest the need to use it early.

Keywords: Tuberculosis; HIV; Rifampin; Drug toxicity; Risk factors; Medication adherence.