BACKGROUND: Previous work established non-inferiority of switching participants who were virologically suppressed from daily oral standard of care to monthly long-acting intramuscular injections of cabotegravir plus rilpivirine over 96 weeks following a cabotegravir plus rilpivirine oral lead-in. Here, we report an evaluation of switching participants from standard of care oral regimens to long-acting cabotegravir plus rilpivirine via direct-to-injection or oral lead-in pathways. METHODS: This study reports the week 124 results of the FLAIR study, an ongoing phase 3, randomised, open-label, multicentre (11 countries) trial. Antiretroviral therapy (ART)-naive participants who were virologically suppressed (HIV-1 RNA <50 copies per mL) during the 20-week induction phase with standard of care were randomly assigned (1:1) to continue the standard of care oral regimen or switch to long-acting cabotegravir plus rilpivirine (283 per group) in the 100-week maintenance phase. Randomisation was stratified by sex at birth and baseline (pre-induction) HIV-1 RNA (<100 000 or ≥100 000 copies per mL). Participants randomly assigned to long-acting therapy at baseline received a cabotegravir (30 mg) plus rilpivirine (25 mg) once daily oral lead-in for at least 4 weeks before first injection and could choose to continue long-acting cabotegravir (400 mg) plus rilpivirine (600 mg) every 4 weeks from week 100 or withdraw. At week 100, participants in the oral comparator ART group, in discussion with the investigator, could elect to switch to long-acting therapy (extension switch population), either direct-to-injection or with a 4 week oral lead-in (oral lead-in group), or withdraw. Week 124 endpoints included plasma HIV-1 RNA 50 or more copies per mL and less than 50 copies per mL (US Food and Drug Administration [FDA] Snapshot), confirmed virological failure (two consecutive HIV-1 RNA ≥200 copies per mL), and safety and tolerability. The study is registered at ClinicalTrials.gov, NCT02938520. FINDINGS: Screening occurred between Oct 27, 2016, and March 24, 2017. At week 100, 232 (92%) of 253 participants transitioned to long-acting cabotegravir plus rilpivirine in the extension phase (111 [48%] in the direct-to-injection group and 121 [52%] in the oral lead-in group; extension switch population). 243 (86%) of the 283 who were randomly assigned to the long-acting therapy group continued the long-acting regimen into the extension phase. One (<1%) participant in each extension switch group had 50 or more HIV-1 RNA copies per mL; 110 (99%) participants in the direct-to-injection group and 113 (93%) participants in the oral lead-in group remained suppressed (HIV-1 RNA <50 copies per mL) at the week 124 Snapshot. The lower suppression rates in the oral lead-in group were driven by non-virological reasons. For participants in the randomly assigned long-acting group, 227 (80%) of 283 participants remained suppressed; at the week 124 Snapshot, 14 (5%) participants had HIV-1 RNA 50 or more copies per mL, including five additional participants since the week 96 analysis. The remaining 42 (15%) participants in the randomly assigned long-acting group had no virological data. Adverse events leading to withdrawal were infrequent, occurring in three (1%) participants in the extension switch population (one in the direct-to-injection group and two in the oral lead-in group) after 24 weeks of cabotegravir plus rilpivirine therapy, and 15 (5%) participants in the randomly assigned long-acting group up to 124 weeks of therapy. No deaths occurred in the extension phase. Overall, cabotegravir plus rilpivirine adverse event type, severity, and frequency were similar across all groups. Injection site reactions were the most common adverse event, occurring after 914 (21%) of 4442 injections in the extension switch population and 3732 (21%) of 17 392 injections in the randomly assigned long-acting group. Injection site reactions were mostly classified as mild-to-moderate in severity and decreased in incidence over time. Four (2%) of 232 participants in the extension switch population and seven (2%) of 283 in the randomly assigned long-acting group withdrew due to injection-related reasons. INTERPRETATION: After 24 weeks of follow-up, switching to long-acting treatment with or without an oral lead-in phase had similar safety, tolerability, and efficacy, supporting future evaluation of the simpler direct-to-injection approach. The week 124 results for participants randomly assigned originally to the long-acting therapy show long-acting cabotegravir plus rilpivirine remains a durable maintenance therapy with a favourable safety profile. FUNDING: ViiV Healthcare and Janssen Research & Development.