Summary Background Longer duration of first-line chemotherapy for patients with metastatic breast cancer is associated with prolonged overall survival and improved progression-free survival. We investigated capecitabine added to maintenance bevacizumab after initial treatment with bevacizumab and docetaxel in this setting. Methods We did this open-label randomised phase 3 trial at 54 hospitals in Brazil, China, Egypt, France, Hong Kong, India, Italy, Poland, Spain, and Turkey. We enrolled patients with HER2-negative measurable metastatic breast cancer; each received three to six cycles of first-line bevacizumab (15 mg/kg) and docetaxel (75–100 mg/m2 ) every 3 weeks. Progression-free patients were randomly assigned with an interactive voice-response system by block (size four) randomisation (1:1) to receive either bevacizumab and capecitabine or bevacizumab only (bevacizumab 15 mg/kg on day 1; capecitabine 1000 mg/m2 twice per day on days 1–14, every 3 weeks) until progression, stratified by oestrogen receptor status (positive vs negative), visceral metastases (present vs absent), response status (stable disease vs response vs non-measurable), and lactate dehydrogenase concentration (≤1·5 vs >1·5 × upper limit of normal). Neither patients nor investigators were masked to allocation. The primary endpoint was progression-free survival (from randomisation) in the intention-to-treat population. This trial is registered with ClinicalTrials.gov , NCT00929240. Findings Between July 16, 2009, and March 7, 2011 (when enrolment was prematurely terminated), 284 patients received initial bevacizumab and docetaxel; 185 (65%) were randomly assigned (91 to bevacizumab and capecitabine versus 94 to bevacizumab only). Progression-free survival was significantly longer in the bevacizumab and capecitabine group than in the bevacizumab only group (median 11·9 months 95% CI 9·8–15·4 vs 4·3 months 3·9–6·8; stratified hazard ratio 0·38 95% CI 0·27–0·55; two-sided log-rank p<0·0001), as was overall survival (median 39·0 months 95% CI 32·3–not reached vs 23·7 months 18·5–31·7; stratified HR 0·43 95% CI 0·26–0·69; two-sided log-rank p=0·0003). Results for time to progression were consistent with those for progression-free survival. 78 (86%) patients in the bevacizumab and capecitabine group and 72 (77%) in the bevacizumab only group had an objective response. Clinical benefit was recorded in 92 (98%) patients in the bevacizumab alone group and 90 (99%) in the bevacizumab and capecitabine group. Mean change from baseline in global health score did not differ significantly between groups. Grade 3 or worse adverse events during the maintenance phase were more common with bevacizumab and capecitabine than with bevacizumab only (45 49% of 91 patients vs 25 27% of 92 patients). The most common grade 3 or worse events were hand–foot syndrome (28 31% in the bevacizumab and capecitabine group vs none in the bevacizumab alone group), hypertension (eight 9% vs three 3%), and proteinuria (three 3% vs four 4%). Serious adverse events were reported by ten (11%) patients in the bevacizumab and capecitabine group and seven (8%) patients in the bevacizumab only group. Interpretation Despite prematurely terminated accrual and the lack of information about post-progression treatment, both progression-free survival and overall survival were significantly improved with bevacizumab and capecitabine compared with bevacizumab alone as maintenance treatment. These results might inform future maintenance trials and current first-line treatment strategies for HER2-negative metastatic breast cancer. Funding F Hoffmann-La Roche.