The efficacy and safety of rifabutin (RBT) and rifampicin (RMP) were compared in 298 patients with newly diagnosed pulmonary tuberculosis. In the initial 8-wk phase, all patients received isoniazid 400 mg/d, ethambutol 1200 mg/d, and pyrazinamide 2 g/d and were randomly allocated to receive either RMP 600 mg/d or RBT 300 mg/d. In the 16-wk continuation phase, patients received intermittent treatment (twice weekly) with isoniazid 600 mg/d, ethambutol 2400 mg/d and either RMP 600 mg/d or RBT 300 mg/d. Two hundred twenty-five (RMP = 118; RBT = 107) patients completed the 24-wk treatment period (evaluable patient population). Bacteriologic conversion rates in the RMP and RBT groups were 87.7 versus 92.0% at Week 8, 99.1 versus 99.0% at Week 12, 93.5 versus 93.8% at Week 24, and 89.8 versus 95.3% at the last valid observation. The mean time to first bacteriologic conversion was 14.1 wk in the RMP group and 14.3 wk in the RBT group. None of these differences was significant. Adverse events were reported by four patients (five events) in the RMP group and six patients (six events) in the RBT group. Those events thought to be associated with RMP were increased SGOT and leucopenia and, with RBT, increased SGOT and thrombocytopenia. Two hundred four patients entered the follow-up phase, and, of these, 95 (RMP = 49; RBT = 46) completed the scheduled 24-mo period. The overall rate of relapse was 3.8% (4/106) for the RMP group and 5.1% (5/98) for the RBT group. These differences were not significant. All relapsed patients, except for two who could not be traced, were successfully retreated. We conclude that the efficacy and tolerability of RBT is equivalent to that of RMP in the treatment of newly diagnosed uncomplicated tuberculosis, and that RBT can be effectively administered in a part-daily, part-intermittent dosage schedule.