Infection with the human immunodeficiency virus (HIV) has changed both the epidemiology and natural history of tuberculosis. Despite a generally good response to effective antituberculous therapy, the prognosis remains poor. The objective of this analysis was to determine the independent predictors of survival in HIV-infected Ugandan adults with smear-positive pulmonary tuberculosis. A total of 191 HIV-infected Ugandan adults with smear-positive pulmonary tuberculosis were enrolled into a clinical trial of chemotherapy for tuberculosis. The subjects received either rifampin, isoniazid, and pyrazinamide for two months, followed by rifampin and isoniazid for six months (n = 101) or streptomycin, thiacetazone, and isoniazid for two months followed by thiacetazone and isoniazid for eight months (n = 90). After standard measurements were made at baseline, the group was followed at regular intervals for a mean of 16 months to determine survival. During the course of follow-up, 82 (43%) of the patients died, six within the first month of therapy. The one-year survival proportion was 68% with an estimated median survival of 26 months and did not differ according to treatment regimen. The hazard for death was biphasic, high early in the course of therapy, and then again after about one year. After controlling for the treatment regimen, four independent predictors of survival were found: anergy to purified protein derivative, atypical chest roentgenogram, previous HIV-related condition, and lymphopenia. In this cohort of Ugandan adults, four simple and inexpensive predictors of survival were found. These factors suggest that the degree of immunosuppression was a major determinant of survival.