Background Antiangiogenic agents have established efficacy in the treatment of metastatic colorectal cancer. We investigated whether bevacizumab could improve disease-free survival in the adjuvant setting after resection of the primary tumour. Methods For the open-label, randomised, controlled QUASAR 2 trial, which was done at 170 hospitals in seven countries, we recruited patients aged 18 years or older with WHO performance status scores of 0 or 1 who had undergone potentially curative surgery for histologically proven stage III or high-risk stage II colorectal cancer. Patients were randomly assigned (1:1) to receive eight 3-week cycles of oral capecitabine alone (1250 mg/m2 twice daily for 14 days followed by a break for 7 days) or the same regimen of oral capecitabine plus 16 cycles of 7·5 mg/kg bevacizumab by intravenous infusion over 90 min on day 1 of each cycle. Randomisation was done by a computer-generated schedule with use of minimisation with a random element stratified by age, disease stage, tumour site, and country. The study was open label and no-one was masked to treatment assignment. The primary endpoint was 3-year disease-free survival, assessed in the intention-to-treat population. Toxic effects were assessed in patients who received at least one dose of randomised treatment. This trial is registered with the ISRCTN registry, number ISRCTN45133151. Findings Between April 25, 2005, and Oct 12, 2010, 1952 eligible patients were enrolled, of whom 1941 had assessable data (968 in the capecitabine alone group and 973 in the capecitabine and bevacizumab group). Median follow-up was 4·92 years (IQR 4·00–5·16). Disease-free survival at 3 years did not differ between the groups (75·4%, 95% CI 72·5–78·0 in the capecitabine and bevacizumab group vs 78·4%, 75·7–80·9 in the capecitabine alone group; hazard ratio 1·06, 95% CI 0·89–1·25, p=0·54). The most common grade 3–4 adverse events were hand–foot syndrome (201 [21%] of 963 in the capecitabine alone group vs 257 [27%] of 959 in the capecitabine and bevacizumab group) and diarrhoea (102 [11%] vs 104 [11%]), and, with the addition of bevacizumab, expected increases were recorded in all-grade hypertension (320 [33%] vs 75 [8%]), proteinuria (197 [21%] vs 49 [5%]), and wound healing problems (30 [3%] vs 17 [2%]). 571 serious adverse events were reported (221 with capecitabine alone and 350 with capecitabine and bevacizumab). Most of these were gastrointestinal (n=245) or cardiovascular (n=169). 23 deaths within 6 months of randomisation were classified as being related to treatment, eight in the capecitabine alone group and 15 in the capecitabine and bevacizumab group. Interpretation The addition of bevacizumab to capecitabine in the adjuvant setting for colorectal cancer yielded no benefit in the treatment of an unselected population and should not be used. Funding Roche.