INTRODUCTION: Mortality from tuberculosis, which should be a rare event, still affects a large portion of the population of developing countries. In this context, Recife, a city in the northeast of Brazil where this study was developed, has the highest tuberculosis mortality rates of the Brazilian capitals. OBJECTIVE: To analyze survival probability and identify risk factors for death from tuberculosis in a cohort of patients living in Recife who started treatment for tuberculosis. METHODOLOGY: A cohort of newly diagnosed TB cases was followed up from the beginning of treatment (in 2001-2003) until June 2007. Survival probability was estimated by Kaplan-Meier method; and Cox Regression analysis was used to identify risk factors. RESULTS: At the end of the follow-up period, the survival probability after beginning TB treatment was 95.9%. Older ages, positivity for HIV and late initial treatment were statistically associated with death from TB in one year follow-up. When the analysis was done considering the total period of follow-up, older ages, positivity serology for HIV, late initial treatment, weight loss, and history of previous treatment remained in the multivariate Cox regression model. CONCLUSION: A more comprehensive analysis, specifically for deaths from tuberculosis as the underlying and non-underlying cause, allowed identification of a greater number of predictive factors that would otherwise not be detected if follow-up had lasted only until the end of treatment. These results can guide feasible interventions for health services aiming to reduce case-fatality from tuberculosis.