Background Phase 3 clinical studies showed non-inferiority of long-acting intramuscular cabotegravir and rilpivirine dosed every 4 weeks to oral antiretroviral therapy. Important phase 2 results of every 8 weeks dosing, and supportive modelling, underpin further evaluation of every 8 weeks dosing in this trial, which has the potential to offer greater convenience. Our objective was to compare the week 48 antiviral efficacy of cabotegravir plus rilpivirine long-acting dosed every 8 weeks with that of every 4 weeks dosing. Methods ATLAS-2M is an ongoing, randomised, multicentre (13 countries; Australia, Argentina, Canada, France, Germany, Italy, Mexico, Russia, South Africa, South Korea, Spain, Sweden, and the USA), open-label, phase 3b, non-inferiority study of cabotegravir plus rilpivirine long-acting maintenance therapy administered intramuscularly every 8 weeks (cabotegravir 600 mg plus rilpivirine 900 mg) or every 4 weeks (cabotegravir 400 mg plus rilpivirine 600 mg) to treatment-experienced adults living with HIV-1. Eligible newly recruited individuals must have received an uninterrupted first or second oral standard-of-care regimen for at least 6 months without virological failure and be aged 18 years or older. Eligible participants from the ATLAS trial, from both the oral standard-of-care and long-acting groups, must have completed the 52-week comparative phase with an ATLAS-2M screening plasma HIV-1 RNA less than 50 copies per mL. Participants were randomly assigned 1:1 to receive cabotegravir plus rilpivirine long-acting every 8 weeks or every 4 weeks. The randomisation schedule was generated by means of the GlaxoSmithKline validated randomisation software RANDALL NG. The primary endpoint at week 48 was HIV-1 RNA >= 50 copies per mL (Snapshot, intention-to-treat exposed), with a non-inferiority margin of 4%. The trial is registered at ClinicalTrials.gov, NCT03299049 and is ongoing. Findings Screening occurred between Oct 27, 2017, and May 31, 2018. Of 1149 individuals screened, 1045 participants were randomised to the every 8 weeks (n=522) or every 4 weeks (n=523) groups; 37% (n=391) transitioned from every 4 weeks cabotegravir plus rilpivirine long-acting in ATLAS. Median participant age was 42 years (IQR 34-50); 27% (n=280) female at birth; 73% (n=763) white race. Cabotegravir plus rilpivirine long-acting every 8 weeks was non-inferior to dosing every 4 weeks (HIV-1 RNA >= 50 copies per mL; 2% vs 1%) with an adjusted treatment difference of 0.8 (95% CI -0.6-2.2). There were eight (2%, every 8 weeks group) and two (<1%, every 4 weeks group) confirmed virological failures (two sequential measures >= 200 copies per mL). For the every 8 weeks group, five (63%) of eight had archived non-nucleoside reverse transcriptase inhibitor resistance-associated mutations to rilpivirine at baseline. The safety profile was similar between dosing groups, with 844 (81%) of 1045 participants having adverse events (excluding injection site reactions); no treatment-related deaths occurred. Interpretation The efficacy and safety profiles of dosing every 8 weeks and dosing every 4 weeks were similar. These results support the use of cabotegravir plus rilpivirine long-acting administered every 2 months as a therapeutic option for people living with HIV-1.