Bevacizumab is approved in combination with chemotherapy for the treatment of ovarian cancer, either in first-line therapy or for patients with recurrent disease not previously treated with the same drug. We aimed to test the value of continuing bevacizumab beyond progression after first-line treatment with the same drug.In our open-label, randomised, phase 3 trial done at 82 sites in four countries, we enrolled women (aged ≥18 years) who had previously received first-line platinum-based therapy including bevacizumab, and had recurrent (≥6 months since last platinum dose), International Federation of Gynaecology and Obstetrics stage IIIB-IV ovarian cancer with an Eastern Cooperative Oncology Group performance status 0-2. Patients were randomly assigned (1:1) to receive a carboplatin-based doublet intravenously (carboplatin area under the concentration curve [AUC] 5 on day 1 plus paclitaxel 175 mg/m2 on day 1, every 21 days; carboplatin AUC 4 on day 1 plus gemcitabine 1000 mg/m2 on days 1 and 8, every 21 days; or carboplatin AUC 5 on day 1 plus pegylated liposomal doxorubicin 30 mg/m2 on day 1, every 28 days), or a carboplatin-based doublet plus bevacizumab (10 mg/kg intravenous every 14 days combined with pegylated liposomal doxorubicin-carboplatin, or 15 mg/kg every 21 days combined with gemcitabine-carboplatin or paclitaxel-carboplatin). Evaluable disease according to RECIST 1.1 guidelines was required before randomisation. Randomisation was done through the trial website with a minimisation procedure, stratified by centre, time of recurrence, performance status, and type of second-line chemotherapy. The primary endpoint was investigator-assessed progression-free survival, analysed on an intention-to-treat basis. Safety was assessed in all participants who received at least one dose. This trial is registered with ClinicalTrials.gov, NCT01802749 and EudraCT 2012-004362-17.Between Dec 6, 2013, and Nov 11, 2016, 406 patients were recruited (203 [50%] assigned to the bevacizumab group and 203 [50%] to the standard chemotherapy group). 130 patients (64%) in the bevacizumab group and 131 (65%) in the standard chemotherapy group had progressed after receiving a last dose of platinum more than 12 months before, and 146 patients (72%) in the bevacizumab group and 147 (72%) in the standard chemotherapy group had progressed after completion of first-line bevacizumab maintenance. 161 participants (79%) progressed in the standard chemotherapy group, as did 143 (70%) in the bevacizumab group. Median progression-free survival was 8·8 months (95% CI 8·4-9·3) in the standard chemotherapy group and 11·8 months (10·8-12·9) in the bevacizumab group (hazard ratio 0·51, 95% CI 0·41-0·65; log-rank p<0·0001). Most common grade 3-4 adverse events were hypertension (20 [10%] in the standard chemotherapy group vs 58 (29%) in the bevacizumab group), neutrophil count decrease (81 [41%] vs 80 [40%]), and platelet count decrease (43 [22%] vs 61 [30%]). 68 patients (33%) died in the standard chemotherapy group and 79 (39%) died in the bevacizumab group; two deaths (1%) in the standard chemotherapy group and one death (<1%) in the bevacizumab group were deemed to be treatment-related.Continuing bevacizumab beyond progression combined with chemotherapy in patients with platinum-sensitive recurrent ovarian cancer improves progression-free survival compared with standard chemotherapy alone and might be considered in clinical practice.Hoffmann-La Roche and Associazione Italiana per la Ricerca sul Cancro.