Background: Randomised phase 3 trials in metastatic breast cancer have shown that combining bevacizumab with either paclitaxel or capecitabine significantly improves progression-free survival and response rate compared with chemotherapy alone but the relative efficacy of bevacizumab plus paclitaxel versus bevacizumab plus capecitabine has not been investigated. We compared the efficacy of the two regimens. Methods: In this open-label, non-inferiority, phase 3 trial, patients with HER2-negative metastatic breast cancer who had received no chemotherapy for advanced disease were randomised (by computer-generated sequence; 1:1 ratio; block size six; stratified by hormone receptor status, country, and menopausal status) to receive either intravenous bevacizumab (10 mg/kg on days 1 and 15) plus intravenous paclitaxel (90 mg/m2 on days 1, 8, and 15) repeated every 4 weeks (paclitaxel group) or intravenous bevacizumab (15 mg/kg on day 1) plus oral capecitabine (1000 mg/m2 twice daily on days 1-14) repeated every 3 weeks (capecitabine group) until disease progression or unacceptable toxic effects. Treatment allocation was not masked because of the differences in routes of administration and cycle lengths. The primary objective was to show non-inferior overall survival with bevacizumab plus capecitabine versus bevacizumab plus paclitaxel. We report results of an interim overall survival analysis, which was planned for after 175 deaths in the per-protocol population. This trial is registered with ClinicalTrials.gov, number NCT00600340. Findings: Between Sept 10, 2008, and Aug 30, 2010, we randomised 564 patients (paclitaxel group n=285; capecitabine group n=279) from 51 centres in 12 countries. The per-protocol population consisted of 533 patients (paclitaxel group n=268; capecitabine group n=265). After median follow-up of 18·6 months (IQR 14·9-24·7), 181 patients in the per-protocol population had died (89 [33%] in the paclitaxel group; 92 [35%] in the capecitabine group). The hazard ratio [HR] for overall survival was 1·04 (97·5% repeated CI -∞ to 1·69; p=0·059); the non-inferiority criterion of the interim analysis (interim α=0·00105) was not met. More patients who received bevacizumab plus paclitaxel had an objective response than did those who received bevacizumab plus capecitabine (125 [44%] of 285 patients vs 76 [27%] of 279; p<0·0001). Similarly, progression-free survival was significantly longer in the paclitaxel group than in the capecitabine group (median progression-free survival 11·0 months [95% CI 10·4-12·9] vs 8·1 months [7·1-9·2]; HR 1·36 [95% CI 1·09-1·68], p=0·0052). The most common adverse events of grade 3 or higher were neutropenia (51 [18%]), peripheral neuropathy (39 [14%]), and leucopenia (20 [7%]) in the paclitaxel group and hand-foot syndrome (44 [16%]), hypertension (16 [6%]), and diarrhoea (15 [5%]) in the capecitabine group. One treatment-related death occurred in the paclitaxel group; no deaths in the capecitabine group were deemed to be treatment-related. Interpretation: In this planned interim analysis, the non-inferiority criterion was not met and overall survival results are inconclusive. Final results are expected in 2014. Progression-free survival was better, and more patients achieved an objective response, with bevacizumab plus paclitaxel than with bevacizumab plus capecitabine. Efficacy results in both groups were consistent with previous reports. Funding: Central European Cooperative Oncology Group; Roche.