Summary Background Immunochemotherapy with rituximab and cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) has become the standard of care for elderly patients with diffuse large B-cell lymphoma. We aimed to ascertain if a dose-dense R-CHOP regimen administered every 2 weeks (R-CHOP14) was superior to the standard 3-week schedule (R-CHOP21). Methods We did a randomised phase 3 trial at 83 centres in four countries. 602 patients aged 60–80 years with untreated diffuse large B-cell lymphoma and at least one adverse prognostic factor (age-adjusted international prognostic index ≥1) were eligible for the study. We randomly allocated individuals to R-CHOP—ie, rituximab (375 mg/m2 ), cyclophosphamide (750 mg/m2 ), doxorubicin (50 mg/m2 ), vincristine (1·4 mg/m2 , up to 2 mg) all on day 1, and prednisone 40 mg/m2 daily for 5 days—administered every 14 days (n=304) or every 21 days (n=298) for eight cycles. We did permuted-block randomisation (block size four, allocation ratio 1:1) stratified by centre and number of adverse prognostic factors. The primary endpoint was event-free survival. Our analysis was of the intention-to-treat population, and we present the final analysis. This study is registered with ClinicalTrials.gov , number NCT00144755. Findings Two patients allocated R-CHOP21 were ineligible for the study and were excluded from analyses. After median follow-up of 56 months (IQR 27–60), 3-year event-free survival was 56% (95% CI 50–62) in the R-CHOP14 group and 60% (55–66) in the R-CHOP21 group (hazard ratio 1·04, 95% CI 0·82–1·31; p=0·7614). Grade 3–4 neutropenia occurred in 224 (74%) of 304 patients allocated R-CHOP14 and 189 (64%) of 296 assigned R-CHOP21, despite increased use of granulocyte colony-stimulating factor in the R-CHOP14 group compared with the R-CHOP21 group. 143 (47%) patients in the R-CHOP14 group received at least one red-blood-cell transfusion versus 93 (31%) in the R-CHOP21 group (p=0·0001). 35 (12%) patients allocated R-CHOP14 received at least one platelet transfusion versus 25 (8%) assigned R-CHOP21 (p=0·2156). 155 (51%) patients who were assigned R-CHOP14 had at least one serious adverse event compared with 140 (47%) who were allocated R-CHOP21. Interpretation In elderly patients with untreated diffuse large B-cell lymphoma and at least one adverse prognostic factor, a 2-week dose-dense R-CHOP regimen did not improve efficacy compared with the 3-week standard schedule. The frequency of toxic side-effects was similar between regimens, but R-CHOP14 was associated with increased need for red-blood-cell transfusion. Funding Groupe d'Etude des Lymphomes de l'Adulte (GELA), Amgen.