We sought to determine the risk of acquired rifamycin resistant (ARR) tuberculosis associated with rifampin- versus rifabutin-based directly observed therapy and to assess the risk factors for relapse of tuberculosis. This observational cohort study included patients with culture-confirmed rifamycin-susceptible tuberculosis reported to the Baltimore City Health Department (Baltimore, MD) during the period of January 1993 through December 2001. Of the 407 patients, 108 (27%) were human immunodeficiency virus (HIV) seropositive, 161 (40%) were HIV seronegative, and 138 (34%) had an unknown serostatus. Three (2.8%) of 108 HIV-seropositive persons had ARR tuberculosis, compared with 0 of 299 persons with negative or unknown HIV serostatus (P=.02). Among HIV-seropositive patients, 3 (3.7%) of 81 who were treated with rifampin and 0 of 27 who were treated with rifabutin had ARR tuberculosis (P=.57). Among HIV-seropositive patients, the only risk factor for recurrent tuberculosis was a low median initial CD4+ T lymphocyte count (51 vs. 138 cells/mm3; P=.02). The median CD4+ T lymphocyte count among patients with ARR tuberculosis was 51 cells/mm3. ARR tuberculosis can occur with rifampin-based regimens, but in this study, the risk was not significantly higher than that for a rifabutin-based regimen.