objective. the objective of this case-control study was to identify prognostic factors for the outcome of pulmonary tuberculosis among 297 patients who were treated between 1994 and 1999 at the federal university of pernambuco clinics hospital, in recife, pernambuco, brazil. methods. the cases were defined as individuals whose treatment ended in death, dropout, or failure. the controls were persons whose treatment ended in cure. the following independent variables were submitted to uni- and multivariate analyses: sex, age, schooling, being a smoker, consuming alcohol, previous treatment for tuberculosis, response to the tuberculin test, hiv serology, resistance to antimicrobial agents, results of direct investigation of alcohol- and acid-resistant bacilli, and treatment approach used. after that, the uni- and multivariate analyses were repeated, including as cases only deaths and individuals with treatment failure. results. four risk factors for tuberculosis treatment failure were found: excessive alcohol intake (odds ratio (or) = 2.58; p = 0.014), co-infection with hiv (or = 3.40; p = 0.028), previous tuberculosis treatment (or = 4.89; p < 0.001), and resistance to two or more antituberculosis drugs (or = 3.49; p = 0.017). in the second multivariate analysis, which excluded dropout cases, no association was found between treatment outcome and excessive alcohol consumption, but the other associations remained. this result suggests a close relationship between alcoholism and treatment dropout. conclusions. in the group that was studied the prognostic factors for failure of pulmonary tuberculosis treatment were interrelated and were of a biological, clinical, and social character. these factors should be identified at the beginning of treatment in order to allow implementation of specific follow-up procedures such as the strategy of directly observed treatment. this would strengthen tuberculosis control at the local level.