Background Cabotegravir and rilpivirine are antiretroviral drugs in development as long-acting injectable formulations. The LATTE-2 study evaluated long-acting cabotegravir plus rilpivirine for maintenance of HIV-1 viral suppression through 96 weeks. Methods In this randomised, phase 2b, open-label study, treatment-naive adults infected with HIV-1 initially received oral cabotegravir 30 mg plus abacavir-lamivudine 600-300 mg once daily. The objective of this study was to select an intramuscular dosing regimen based on a comparison of the antiviral activity, tolerability, and safety of the two intramuscular dosing regimens relative to oral cabotegravir plus abacavir-lamivudine. After a 20-week induction period on oral cabotegravir plus abacavir-lamivudine, patients with viral suppression (plasma HIV-1 RNA <50 copies per mL) were randomly assigned (2:2:1) to intramuscular long-acting cabotegravir plus rilpivirine at 4-week intervals (long-acting cabotegravir 400 mg plus rilpivirine 600 mg; two 2 mL injections) or 8-week intervals (long-acting cabotegravir 600 mg plus rilpivirine 900 mg; two 3 mL injections) or continued oral cabotegravir plus abacavir-lamivudine. Randomisation was computer-generated with stratification by HIV-1 RNA (<50 copies per mL, yes or no) during the first 12 weeks of the induction period. The primary endpoints were the proportion of patients with viral suppression at week 32 (as defined by the US Food and Drug Administration snapshot algorithm), protocol-defined virological failures, and safety events through 96 weeks. All randomly assigned patients who received at least one dose of study drug during the maintenance period were included in the primary efficacy and safety analyses. The primary analysis used a Bayesian approach to evaluate the hypothesis that the proportion with viral suppression for each long-acting regimen is not worse than the oral regimen proportion by more than 10% (denoted comparable) according to a prespecified decision rule (ie, posterior probability for comparability >90%). Difference in proportions and associated 95% CIs were supportive to the primary analysis. The trial is registered at ClinicalTrials.gov, number NCT02120352. Findings Among 309 enrolled patients, 286 were randomly assigned to the maintenance period ( 115 to each of the 4-week and 8-week groups and 56 to the oral treatment group). This study is currently ongoing. At 32 weeks following randomisation, both long-acting regimens met primary criteria for comparability in viral suppression relative to the oral comparator group. Viral suppression was maintained at 32 weeks in 51 (91%) of 56 patients in the oral treatment group, 108 (94%) of 115 patients in the 4-week group (difference 2.8% [95% CI -5.8 to 11.5] vs oral treatment), and 109 (95%) of 115 patients in the 8-week group (difference 3.7% [-4.8 to 12.2] vs oral treatment). At week 96, viral suppression was maintained in 47 (84%) of 56 patients receiving oral treatment, 100 (87%) of 115 patients in the 4-week group, and 108 (94%) of 115 patients in the 8-week group. Three patients (1%) experienced protocol-defined virological failure (two in the 8-week group; one in the oral treatment group). Injection-site reactions were mild (3648 [84%] of 4360 injections) or moderate (673 [15%] of 4360 injections) in intensity and rarely resulted in discontinuation (two [<1%] of 230 patients); injection-site pain was reported most frequently. Serious adverse events during maintenance were reported in 22 (10%) of 230 patients in the intramuscular groups (4-week and 8-week groups) and seven (13%) of 56 patients in the oral treatment group; none were drug related. Interpretation The two-drug combination of all-injectable, long-acting cabotegravir plus rilpivirine every 4 weeks or every 8 weeks was as effective as daily three-drug oral therapy at maintaining HIV-1 viral suppression through 96 weeks and was well accepted and tolerated.