The aim of this double-blind, randomized, placebo-controlled, 12-week study was to assess the efficacy of rifabutin (450 or 600 mg/d) in the treatment of disseminated nontuberculous mycobacterial infection in patients with AIDS. Companion drugs in both arms of the study were ethambutol, clofazimine, and isoniazid. Because of low accrual, the study was prematurely terminated when a total of 382 patients had been enrolled, of which 200 were eligible (i.e., their specimens were culture-positive for Mycobacterium avium complex [MAC] or Mycobacterium xenopi at baseline) and 102 were evaluable (i.e., they were eligible, were treated for a minimum of 6 weeks, and had at least one culture assessment after baseline). The original protocol called for a total of 220 evaluable patients. At week 12, rifabutin treatment was associated with higher, although nonsignificant, rates of bacteriologic conversion than was the placebo arm, with regard to both the eligible patients (25% vs. 18%) and the evaluable patients (45% vs. 38%). Corresponding median times to culture conversion were 42 vs. 63 days (eligible patients) and 43 vs, 69 days (evaluable patients). No significant difference was observed in clinical improvement, mortality, or toxicity between the two treatment arms. The addition of rifabutin to a triple-drug regimen may contribute to the clearance of disseminated MAC infection in patients with AIDS, without causing additional toxicity.