Objective: Compare the effectiveness, tolerability, and safety of 3 months of weekly rifapentine and isoniazid under direct observation (3HP) versus 9 months of daily isoniazid (9H) in HIV-infected persons. Design: Prospective, randomized, and open-label noninferiority trial. Setting: The United States, Brazil, Spain, Peru, Canada, and Hong Kong. Participants: HIV-infected persons who were tuberculin skin test positive or close contacts of tuberculosis cases. Intervention: 3HP versus 9H. Main outcome measures: The effectiveness endpoint was tuberculosis; the noninferiority margin was 0.75%. The tolerability endpoint was treatment completion; the safety endpoint was drug discontinuation because of adverse drug reaction. Results: Median baseline CD4 + cell counts were 495 (IQR 389-675) and 538 (IQR 418-729) cells/μl in the 3HP and 9H arms, respectively (P = 0.09). In the modified intention-to-treat analysis, there were two tuberculosis cases among 206 persons [517 person-years (p-y) of follow-up] in the 3HP arm (0.39 per 100 p-y) and six tuberculosis cases among 193 persons (481 p-y of follow-up) in the 9H arm (1.25 per 100 p-y). Cumulative tuberculosis rates were 1.01 versus 3.50% in the 3HP and 9H arms, respectively (rate difference: -2.49%; upper bound of the 95% confidence interval of the difference: 0.60%). Treatment completion was higher with 3HP (89%) than 9H (64%) (P < 0.001), and drug discontinuation because of an adverse drug reaction was similar (3 vs. 4%; P = 0.79) in 3HP and 9H, respectively. Conclusion: Among HIV-infected persons with median CD4 + cell count of approximately 500 cells/μl, 3HP was as effective and safe for treatment of latent Mycobacterium tuberculosis infection as 9H, and better tolerated.