STREAM stage 2 showed that two bedaquiline-containing regimens (a 9-month all-oral regimen and a 6-month regimen with 8 weeks of aminoglycoside) had superior efficacy to a 9-month injectable-containing regimen for rifampicin-resistant tuberculosis up to 76 weeks after randomisation. Our objective in this follow-up analysis was to assess the durability of efficacy and safety, including mortality, at 132 weeks. We report the long-term outcomes from STREAM stage 2, a randomised, phase 3 non-inferiority (10% margin) trial in participants (aged ≥15 years) with rifampicin-resistant tuberculosis without fluoroquinolone or aminoglycoside resistance at 13 clinical sites in seven countries (Ethiopia, Georgia, India, Moldova, Mongolia, South Africa, and Uganda). Participants were randomly assigned 1:2:2:2 (via permuted blocks and stratified by site and HIV status plus CD4 cell count) to the 2011 WHO long regimen (terminated early), a 9-month control regimen, a 9-month oral regimen with bedaquiline (primary comparison), or a 6-month regimen with bedaquiline and 8 weeks of an injectable antituberculous drug. Participants and clinicians were aware of treatment-group assignments, but laboratory staff were masked. The primary outcome, reported previously, was favourable status (negative cultures for Mycobacterium tuberculosis without a preceding unfavourable outcome; any death, bacteriological failure or recurrence, and major treatment change were considered unfavourable) at week 76. Here we report efficacy outcomes at week 132, analysed in the modified intention-to-treat (mITT) population. Safety assessments continued to 132 weeks and were in all participants who received at least one dose of the study regimen. All comparisons used concurrently randomised participants. This trial is registered on ISRCTN (ISRCTN18148631) and is now completed. Between March 28, 2016, and Jan 28, 2020, 588 participants were randomly assigned to the long (n=32), control (n=202), oral (n=211), or 6-month (n=143) treatment regimens; 352 (60%) were male and 236 (40%) were female. Of the 556 participants on the three shorter regimens, 517 were included in the mITT population (187 in control group, 196 in oral group, and 134 in 6-month group) and 465 in the per-protocol analyses. Six additional participants had an unfavourable outcome that occurred between week 76 and the end of efficacy follow-up (one in control group, four in oral group, one in 6-month group). In the mITT population, the proportion of patients with an unfavourable outcome at the end of follow-up was 19·6% (95% CI 14·3 to 24·9) in the oral group and 29·3% (23·3 to 36·5) in the control group (-9·7 percentage points difference [95% CI -18·7 to -1·8]; psuperiority=0·024). An estimated 9·8% (95% CI 4·6 to 14·9) of participants on the 6-month regimen had an unfavourable outcome, which was significantly lower than for those concurrently on the control regimen (32·5% [23·7 to 40·2]; psuperiority<0·0001) or the oral regimen (23·8% [16·9 to 31·1]; psuperiority=0·013). Few serious or severe adverse events were reported after week 76, with no indication of a difference between the regimens. At week 132, treatment-emergent hearing loss was recorded in significantly fewer participants on the oral regimen (7/205; 3%) than the control regimen (16/198; 8%; p=0.041); there was no significant difference in severe hearing loss between the oral regimen (6/139; 4%) and the 6-month regimen (5/143; 4%; p=0·72). Death rates were low: 1·01 (95% CI 0·48 to 2·12) per 100 person-years in participants allocated to bedaquiline (ie, oral and 6-month regimen, n=287) compared with 1·52 (0·63 to 3·66) in participants on the control regimen (n=140; p=0·49). Both of the bedaquiline-containing regimens maintained superiority to the control regimen, without evidence of increased mortality, providing two additional evidence-based treatment options for patients; previous mortality concerns for bedaquiline were not substantiated. US Agency for International Development and Janssen Research & Development.