Summary Background The non-nucleoside reverse transcriptase inhibitor (NNRTI), rilpivirine (TMC278; Tibotec Pharmaceuticals, County Cork, Ireland), had equivalent sustained efficacy to efavirenz in a phase 2b trial in treatment-naive patients infected with HIV-1, but fewer adverse events. We aimed to assess non-inferiority of rilpivirine to efavirenz in a phase 3 trial with common background nucleoside or nucleotide reverse transcriptase inhibitors (NtRTIs). Methods We undertook a 96-week, phase 3, randomised, double-blind, double-dummy, non-inferiority trial in 98 hospitals or medical centres in 21 countries. We enrolled adults (≥18 years) not previously given antiretroviral therapy and with a screening plasma viral load of 5000 copies per mL or more and viral sensitivity to background N(t)RTIs. We randomly allocated patients (1:1) using a computer-generated interactive web-response system to receive oral rilpivirine 25 mg once daily or efavirenz 600 mg once daily; all patients received an investigator-selected regimen of background N(t)RTIs (tenofovir-disoproxil-fumarate plus emtricitabine, zidovudine plus lamivudine, or abacavir plus lamivudine). The primary outcome was non-inferiority (12% margin on logistic regression analysis) at 48 weeks in terms of confirmed response (viral load <50 copies per mL, defined by the intent-to-treat time to loss of virologic response TLOVR algorithm) in all patients who received at least one dose of study drug. This study is registered with ClinicalTrials.gov , number NCT00543725. Findings From May 22, 2008, we screened 947 patients and enrolled 340 to each group. 86% of patients (291 of 340) who received at least one dose of rilpivirine responded, compared with 82% of patients (276 of 338) who received at least one dose of efavirenz (difference 3·5% 95% CI −1·7 to 8·8; pnon-inferiority <0·0001). Increases in CD4 cell counts were much the same between groups. 7% of patients (24 of 340) receiving rilpivirine had a virological failure compared with 5% of patients (18 of 338) receiving efavirenz. 4% of patients (15) in the rilpivirine group and 7% (25) in the efavirenz group discontinued treatment due to adverse events. Grade 2–4 treatment-related adverse events were less common with rilpivirine (16% 54 patients) than they were with efavirenz (31% 104; p<0·0001), as were rash and dizziness (p<0·0001 for both) and increases in lipid levels were significantly lower with rilpivirine than they were with efavirenz (p<0·0001). Interpretation Despite a slightly increased incidence of virological failures, a favourable safety profile and non-inferior efficacy compared with efavirenz means that rilpivirine could be a new treatment option for treatment-naive patients infected with HIV-1. Funding Tibotec.