BACKGROUND: Rifapentine has a long half-life in serum, which suggests a possible treatment once a week for tuberculosis. We aimed to compare rifapentine and isoniazid once a week with rifampicin and isoniazid twice a week. METHODS: We did a randomised, multicentre, open-label trial in the USA and Canada of HIV-negative people with drug-susceptible pulmonary tuberculosis who had completed 2 months of a 6-month treatment regimen. We randomly allocated patients directly observed treatment with either 600 mg rifapentine plus 900 mg isoniazid once a week or 600 mg rifampicin plus 900 mg isoniazid twice a week. Primary outcome was failure/relapse. Analysis was by intention to treat. FINDINGS: 1004 patients were enrolled (502 per treatment group). 928 successfully completed treatment, and 803 completed the 2-year 4-month study. Crude rates of failure/relapse were 46/502 (9.2%) in those on rifapentine once a week, and 28/502 (5.6%) in those given rifampicin twice a week (relative risk 1.64, 95% CI 1.04-2.58, p=0.04). By proportional hazards regression, five characteristics were independently associated with increased risk of failure/relapse: sputum culture positive at 2 months (hazard ratio 2.8, 95% CI 1.7-4.6); cavitation on chest radiography (3.0, 1.6-5.9); being underweight (3.0, 1.8-4.9); bilateral pulmonary involvement (1.8, 1.0-3.1); and being a non-Hispanic white person (1.8, 1.1-3.0). Adjustment for imbalances in 2-month culture and cavitation diminished the association of treatment group with outcome (1.34; 0.83-2.18; p=0.23). Of participants without cavitation, rates of failure/relapse were 6/210 (2.9%) in the once a week group and 6/241 (2.5%) in the twice a week group (relative risk 1.15; 95% CI 0.38-3.50; p=0.81). Rates of adverse events and death were similar in the two treatment groups. INTERPRETATION: Rifapentine once a week is safe and effective for treatment of pulmonary tuberculosis in HIV-negative people without cavitation on chest radiography. Clinical, radiographic, and microbiological data help to identify patients with tuberculosis who are at increased risk of failure or relapse when treated with either regimen.